Lickety help

A small software company in Texas that makes Windows programs you buy once and browser tools that are free.

The whole catalogue runs on your own machine. There is no account to create, nothing to sign in to, and no server doing the work — which is why there is nothing to subscribe to.

You pay one price, one time, and the software is yours.

There is no renewal, no expiry, and no second invoice. It is not a year's access or a licence that lapses — it is the program, kept.

We can promise that honestly because there is no server behind the software. Nothing costs us money every month that you use it, so nothing has to be charged to you every month either.

One person, up to three of their own machines.

A desktop and a laptop is the normal case, and a third is there so you do not have to think about it. The licence is for a person, not a device — if all three machines are yours and you are the one using them, you are inside the policy.

What it does not cover is one purchase shared across a team or a family who each use it separately. That is a different thing and we would rather you asked than guessed.

No.

Machines die, get rebuilt, and get replaced, and none of that is a reason to pay twice. Install it on the new one.

If you ever find yourself blocked, email support@greatscottlabs.co and it gets sorted out the same day. There is deliberate room in the three-machine policy for exactly this, because a rule that punishes a hard drive failure is a rule written for the wrong person.

Yes — every update to the version you bought, for as long as that version exists. Fixes and new features both.

There is no maintenance plan and no support contract. If you bought version 1, every version 1 release is yours.

Eventually, and it works the way Office used to.

A major new version is expected roughly every two and a half years. That one is a new purchase — but existing owners get at least half off, and, more importantly, **the version you already own keeps working**. It does not stop, degrade, or start asking.

So the honest version of "buy it once" is: you own what you bought, permanently, and if a much later major version is worth having you can buy it at a discount. Staying where you are is always a real option.

Yes, for the paid Windows software. It is the same program, not a crippled demo: a generous run of words to start with, then a weekly allowance.

It does not expire, and it does not nag. If the allowance is enough for you, keep using it.

The browser tools are a different thing — those are free outright, with no allowance and nothing to buy.

It keeps working.

This is the question worth asking of any small company, and the answer here is structural rather than reassuring words. The software does not check in with us, does not need a licence server, and does not need an internet connection to run. There is nothing for anyone to switch off.

A cloud product cannot make that promise at any price, because the moment the company goes the service goes. Something you bought in 2026 still runs in 2036 whether or not we are still shipping.

No. No telemetry, no usage reporting, no crash analytics, and no "help us improve the product" option switched on quietly.

You do not have to take that on trust. Disconnect from the network and use it — it works exactly the same, because nothing it does requires a connection. Or watch the connections it makes and see that there are none.

The browser tools are the same: the file you drop on one is processed by code already on your machine and is never uploaded.

Payments go through Lemon Squeezy, which acts as the merchant of record. That means they take the payment, handle VAT and sales tax for your country or state, and issue the receipt.

We never see or store your card number. If you need an invoice with your company's details on it, that comes from Lemon Squeezy and support@greatscottlabs.co can point you at it.

They are free, and the catch is that we hope you will look at the paid software afterwards. That is the entire business model.

No ads, no sign-up, no watermark, no file-size limit, no "upgrade to unlock". Each one is a single HTML file you can download and keep, and a downloaded copy works with no internet at all.

The one honest limitation: a downloaded copy is frozen. If a tool gets a fix later, your saved file will not have it.

Email support@greatscottlabs.co.

There is no ticket system and no chatbot queue. The person who answers is the person who wrote the software. Most mail is answered the same day; the slowest it gets is a couple of days.

Telling us what you did, what happened, and what you expected instead is usually enough. A screenshot is worth more than a careful description — and there is a free tool for that.

Yes, and both get read.

Bug reports: what you did, what happened, what you expected, and the version number — it is in Help → About in the Windows programs.

Feature requests get answered too, including the "no, and here is why" ones. A request that explains the problem you are stuck on is worth more than one that specifies the solution, because the solution is often simpler than what you had in mind.

Because the moment there is a server, everything else on this page changes.

A server has to be paid for monthly, which means you have to be charged monthly. It can have an outage, a breach, or a change of ownership. It is the thing that turns software you own into access you rent.

Running on your machine is not a limitation we are apologising for. It is the decision everything else here follows from.

Lickety lives in your system tray. Press your hotkey to dictate into whatever window has focus:

• HOLD the hotkey and talk, then release — or
• TAP once to start, tap again to stop.

Both work at all times; no mode to choose. A small overlay near your cursor shows Listening / Transcribing / Inserted.

You don't have to use the keyboard at all. If you turn on the always-visible status bar, you can CLICK it to start and click again to stop.

Press ESCAPE at any point while it's listening to throw the recording away — nothing is transcribed and nothing is typed.

PART-NUMBER AND CODE PREFIXES — empty by default. List the letters your part numbers or codes start with, and two things the recogniser gets wrong are put right for them.

A SPLIT LETTER SUFFIX IS JOINED BACK ON. Say "KW-110QPD" and the recogniser may write "KW-110, QPD"; say "TWM-852XL" and it may write "TWM-852 XL". With KW and TWM listed, both come out as one code. Only one to four CAPITAL letters are joined, never a lone I or A — "the KW-110 I ordered" is a sentence — and never a lower-case word.

A DASH BETWEEN DIGITS IS REMOVED. If you dictate codes like F554201 and they arrive as F55-4201, list F55.

You say: "F55 4201" You get: F55-4201 (the pause becomes a dash) With F55 listed: F554201

THE DASH COMES FROM THE RECOGNISER, not from Lickety — saying the code faster usually avoids it, which is not a reasonable thing to ask of you. This is the only rule in the app that DELETES a character the recogniser produced, which is why it does nothing until you name a prefix.

A DIGIT HAS TO FOLLOW THE PREFIX. That one condition is what keeps ordinary hyphenated codes safe: F-150, T-279 and COVID-19 all have something other than a digit after the letters, so none of them is ever touched — even if you list F or T as a prefix.

Formatting → Number formatting, under the trigger-phrase switch.

CAPITALIZE AFTER A DASH OR BULLET — on by default. When you type a separator and then start dictating, the first word gets a capital.

You type: T-279 - You say: "the restore dialog needs a second confirmation" You get: T-279 - The restore dialog needs a second confirmation

IT FIRES AFTER A DASH, A BULLET, A "1)" OR "1.", OR A TABLE PIPE. Nobody types one of those by accident, so the separator is treated as deliberate — you are starting a new item, and a new item starts with a capital.

TURN IT OFF IF YOU OFTEN DICTATE THE SECOND HALF OF A SENTENCE that already contains a dash. "The result — " followed by "and this is the part that matters" is one sentence, and the capital is wrong there.

Settings → Options.

By default Lickety's indicator only appears while it's doing something. You can make it stay on screen instead: Settings → "Floating status bar".

Once it's visible, it's more than an indicator:

• CLICK IT to start dictating. Click it again to stop.
• DRAG IT anywhere you like. It remembers where you put it, per application — so it can sit in one corner for Word and a different one for Excel, and it follows each app to whichever monitor that app is on.

Clicking is a full alternative to the hotkey. If you'd rather not learn a key combination, or you're somewhere a hotkey doesn't reach, the bar is enough on its own.

Clicking it never takes focus away from what you're typing in, so your text still goes to the right place.

RECENTER ABOVE THE TASKBAR EACH TIME LICKETY STARTS — on by default, and it applies to the DESKTOP position only. Each time the app starts, the bar returns to the centre above the taskbar instead of wherever it was last dragged. That is there so it cannot come back stranded behind desktop icons, or on a monitor you have since unplugged, with no obvious way to find it.

You can still drag it anywhere you like while the app is running, and Word, Excel and every other application keep their own remembered spots either way — this switch never touches those. Turn it off if you always park the desktop bar somewhere particular and want it to stay there between sessions.

Started one by accident? Press Escape — see below.

SHORT VERSION: since version 589 every Lickety hotkey works in a FULL-SCREEN Remote Desktop session — Ctrl+NumPad 1, 2 and 3, a hotkey with a letter in it, and Escape to cancel. Right after the session takes the screen there is a moment, about half a second, before they do; a hotkey pressed in that moment is simply missed, so press it again.

WHY THERE WAS A PROBLEM. When the Remote Desktop window is full screen, the client watches the keyboard itself so shortcuts reach the remote machine instead of yours — that's what makes Alt-Tab switch remote windows. Windows lets the most recent watcher see each key first, and the client passes Ctrl, Shift and Alt on but keeps every other key. Lickety, which runs on YOUR machine, used to start watching once, when it started, so it was behind the client: it saw the Ctrl of Ctrl+NumPad 1 and never the 1. That is why a hotkey made only of modifiers, like Ctrl+Alt, always worked.

WHAT LICKETY DOES NOW. While a Remote Desktop window is in front, Lickety puts its keyboard watcher back in front of the client's — half a second after the window comes to the front or goes full screen, and every ten seconds after that. Nothing changes while a Remote Desktop window isn't in front. Lickety still never keeps a key for itself: a hotkey reaches Lickety AND the remote session, exactly as it does in a windowed session.

THE WINDOWS KEY is still not allowed in a hotkey; Remote Desktop forwards it to the remote machine, and Lickety won't let you set one.

IF A HOTKEY STILL DOES NOTHING in full screen, the status bar works regardless — a click never touches the keyboard — and the client setting below sends every shortcut to your own machine: Show Options → Local Resources → Keyboard → 'Apply Windows key combinations' → 'On this computer'. The cost is that Alt-Tab then switches your LOCAL windows rather than the remote ones.

TRANSCRIBE turns a recording you already have into a text file. Point it at an audio or video file and Lickety writes a transcript beside it.

THIS IS NOT DICTATION, AND THE DIFFERENCE MATTERS. When you dictate, Lickety is helping you write — it fixes your names, obeys "period" and "new line", and tidies as it goes. A recording of other people is EVIDENCE. If you are pulling a quote out of a training session, a quote that was quietly corrected is worse than one with a typo, because nothing on the page tells you it was touched.

So none of your formatting rules run here. Nothing anybody says on the recording is treated as a command. Your Vocabulary, Replacements, capitalisation, dates, numbers and lists are all switched off — what was said is what you get.

THE ONE THING THAT IS REMOVED IS TEXT NOBODY SAID. The recogniser sometimes writes a sentence into a stretch of silence — "Okay.", "Thank you.", "Thanks for watching" — because it was trained on captioned video and that is what usually follows silence there. Those are taken out, and only when the audio underneath them is far quieter than the speech around it. Anything actually spoken stays, including a quiet afterthought at the end. Every removal is written to the dictation log with the levels it compared and the size of what was taken, so it can be checked afterwards. The words themselves go to the diagnostic log, and only if you have turned that on.

ONE MORE CASE, FOR THE SAME REASON. The recogniser listens in 30-second stretches. A word being spoken at the instant one stretch ends is heard by both, and can be written twice. When the sound runs straight through that instant — one word, not the same word said twice with a breath between — the second copy is taken out, and logged the same way. Dictation does this too.

HOW LONG IT TAKES, roughly, on all your cores: about 4 minutes per hour of audio with the base model, and about 15 with the small one. The screen shows an estimate once you have picked a file. **The rest of Lickety keeps working while it runs**, and you can leave the Transcribe screen, or close the window, and come back — it does not stop, and the progress and Cancel are still there when you return. If the window is closed when it finishes, a notice from the tray icon says so (or the transcript opens, if you ticked that). Exiting Lickety while it runs asks first, because the transcript is only written at the end.

WHERE THE TRANSCRIPT GOES IF IT CANNOT GO BESIDE THE RECORDING. Some folders cannot be written to — a disc, a read-only network share — and a recording sitting in a temporary folder is in a place Windows cleans up. Lickety checks this BEFORE it starts, and puts the transcript in your Documents folder instead — the log says so. If saving fails at the end (a full disk, a drive that went away), the transcript is saved in Documents rather than lost.

CANCEL stops the run and nothing is written. The line under the progress bar reads Cancelled. It can take a few seconds to stop while the model is in the middle of a stretch of audio.

HOW LONG A RECORDING CAN BE. Up to 12 hours. The whole recording is held in memory while it is transcribed, so a longer one is refused with a message — split it into shorter files.

WHICH MODEL. Base is roughly three times faster; small is more accurate on unfamiliar names and difficult audio. This is the one place Lickety asks you per job rather than using your dictation setting, because the wait is yours to trade.

THE TWO OPTIONAL EXTRAS, BOTH OFF UNLESS YOU TURN THEM ON.

USE MY VOCABULARY tells the recogniser which words to expect. It helps with names it would otherwise mangle — and it BIASES the recogniser, so it can occasionally insert a term nobody said. On your own dictation that trade is usually worth it. On somebody else's recording, think about it first.

APPLY MY REPLACEMENTS runs your replacement list over the finished text, so a company name still comes out spelled your way.

Whatever you turn on is written into the top of the transcript, so the file always says what was done to it.

ONLY PART OF IT. Fill in FROM, or TO, or both, as mm:ss — 5:00 is five minutes, 0:05 is five seconds. Leave one blank and it means the start or the end, so "the first five minutes" is one box. A bare number is refused on purpose: 5 means five seconds to one person and five minutes to the next, and guessing wrong wastes the whole run. When you transcribe part of a file the transcript says so, in capitals, because a transcript that covers two minutes of an hour and does not mention it is a document that lies about itself.

THE FILE NAME FOLLOWS THE RECORDING. Choose a different file and the transcript is named after it, with a .txt extension, in whatever folder the box already points at — so a name left over from the last run cannot quietly attach itself to a different video. Type the box afterwards if you want something else. {model} {cores} {extras} {date} {time} {from} {to} add to it, landing before the extension, so six runs over one recording do not all land as "(4)". An existing file is never overwritten. The FOLDER button sets where it goes and leaves the name alone.

WHICH FILES IT OPENS. WAV, MP3, M4A, AAC, FLAC, WMA, AIFF, and the audio inside MP4, MOV, MKV, WMV, AVI and most screen recordings. There is nothing to install and nothing to convert first — Lickety uses the audio and video support already built into Windows, so whatever your recorder, phone or meeting app produced should just open.

IF A FILE WILL NOT OPEN, it is usually damaged, still being written, or in an unusual format. Saving or exporting it as MP4, MP3 or WAV almost always fixes it, and the message on screen will say so.

WHAT THIS DOES NOT DO. No speaker labels — it will not tell you who said what. No timestamps in the text. Transcripts do NOT go into your Transcripts screen and do NOT count toward your Stats, because those measure your dictation and an hour of somebody else's talk would swamp them.

TURNING A RAW TRANSCRIPT INTO A READABLE DOCUMENT. A transcript is written as evidence — no shaping, no formatting, sentences often split where the recogniser cut the audio. That is deliberate and it is not going to change: a quote that was quietly corrected is worse than one with a typo. If you want a readable document, click DOCUMENT CLEANUP above the log. It gives you two prompts to paste into whatever AI you already use, and a list of one-line changes to make them do something slightly different. Lickety does not upload anything — it copies text to your clipboard and what happens next is your decision.

A NOTE ON YOUTUBE AND OTHER ONLINE VIDEO. Lickety does not download anything from the internet — it never has and this screen does not change that. If you have a recording of a video on your PC, transcribe it here like any other file.

TAP THE COMMAND HOTKEY (Ctrl+Numpad 5 unless you changed it), say a command, tap it again — or hold it while you speak and let go. It works exactly like the dictation hotkey. The words are matched against your command list and the action runs — a folder opens in Explorer, a file opens, a program starts, or a web page opens. NOTHING IS TYPED into the window you are in, whether the command matched or not. If you press the command hotkey while a dictation is listening, or the dictation hotkey while a command is, the second press is ignored — the two never turn into each other.

"OPEN" PLUS ANY PROGRAM ON YOUR START MENU NEEDS NO SETUP. "Open Word", "open Notepad", "open Calculator" — the names are read when Lickety starts, from your Start Menu's shortcuts and from Windows' own list of installed apps, which is where apps from the Microsoft Store are. Installed something since? Click Refresh Start Menu on the Commands screen. Not sure what an app is called? Click App list there.

"OPEN" PLUS ANY OF LICKETY'S OWN SCREENS WORKS TOO. "Open Notes", "open Transcribe", "open Formatting", "open Replacements" — every screen in the list on the left, by the name it has there, brings the Lickety window up on that screen from wherever you are. "Open command" and "open note" work as well as the plural, "quick text" as two words, and "open Lickety settings" or "open the Notes screen" if you like. A row of your own with the same words still wins. ONE THING TO KNOW: "open settings" opens Lickety's Settings, not Windows'. If you want Windows Settings by voice, add a row: Say "open Windows settings", Do = Open URL, Target = ms-settings:

ALSO BUILT IN, NO SETUP: • "WHAT CAN I SAY?" (or "show commands") opens a list of everything you can say, your own commands included. The same list is behind the WHAT CAN I SAY? button on the Commands screen. • "CREATE A NOTE" (or "create note", "create a new note", "new note", "make a note") brings Lickety up on Notes with a blank note open and the cursor in its title — "open Notes" and the New note button in one breath. If the recogniser writes "node" for "note", that works too. • YOUR FOLDERS: "open downloads", "open documents", "open desktop", "open pictures", "open music", "open videos" — wherever Windows keeps them, OneDrive included. • THE WEB: "search for …", "search the web for …", "google …" open Google on the results; "search YouTube for …" opens YouTube's. Everything after the first words is the search. • FILL-IN COMMANDS OF YOUR OWN: put {} in a row's Target and whatever you say after the row's words goes there. Say "search Amazon for", Do = Open URL, Target = https://www.amazon.com/s?k={} — then "search Amazon for USB cables" opens that search. It is the only kind of row that takes extra words; every other row is its whole phrase. If the recogniser writes "for" as "4" ("search Amazon 4 …"), it is still your row. • TWO OR THREE AT ONCE: join them with "and" — "open music and open videos" opens both. "In" is taken as "and" too, because the recogniser often hears it that way ("open music, in open videos"). Each part has to be a whole command on its own ("open music and videos" does nothing), and a phrase that is already a command — a fill-in like "search for salt and pepper" — is never split. • "PLEASE" at the start or end is fine: "open Notes, please". As always, a row of your own with the same words wins over any of these. Whatever a command opens is brought to the front, even when Windows would have left it behind your window or on the taskbar.

YOUR OWN COMMANDS live on the Commands screen under Tools: one row each — what to say, what to do (open folder, open file, open app, open URL), and the target — %windir%, %LOCALAPPDATA% and the other Windows variables work in it. Put more than one way of saying it in the Say column separated by | — the recogniser sometimes writes "GSL" as "G S L", and that form is understood; "gee ess ell" you add yourself.

WHEN IT DOES NOT MATCH, the floating bar says what was heard, and nothing runs — a wrong folder opening is worse than nothing opening. Open log on the Commands screen lists every press, hits and misses, so you can see exactly what to add to a row.

YOUR VOCABULARY AND YOUR REPLACEMENTS WORK HERE TOO. Vocabulary shapes what is heard on a command press exactly as it does for dictation. Replacements are tried when what was heard matches no command: if "open Claude context" keeps arriving as "open clon context", a Replacement that turns "open clon" into "open Claude" makes the command work, and so does adding the misheard words to the row after a |. A command that matched as heard is never changed by a Replacement.

WHERE THINGS ARE. Your list is Commands.json beside your settings file, backed up on every change; the log is commands.log in the same folder, kept for 90 days. Open log on the Commands screen opens it in Notepad.

ONE ROW IS ONE COMMAND. Click into the empty row at the bottom to add one; a row saves the moment you leave it, and the file beside your settings is backed up five deep on every change.

SAY — the words. "Open GSL folder". Capitals and punctuation do not matter. More than one way of saying it goes in the same box separated by |, which means OR: "open GSL folder | open the GSL folder" is one command you can say either way. Spelled-out letters are understood on their own — say "G S L" and it is read as "GSL".

DO — what happens. OPEN FOLDER shows the folder in Explorer. OPEN FILE opens a file — a document, a spreadsheet, a PDF, a picture — in the program that normally opens that kind of file. OPEN APP starts a program, or a Windows tool such as services.msc. OPEN URL opens a web address in your default browser.

OPEN SEVERAL THINGS WITH ONE COMMAND: give a row more than one target, separated by a semicolon and a space, which means AND. All the targets in a row are the same kind, the one chosen under Do. The two marks are never swapped: | goes in Say and means OR, ; goes in Target and means AND. Open app: "Word; Excel" or "C:\Tools\App1.exe; C:\Tools\App2.exe". A plain name like Word is found in the Start Menu, the same way "open Word" works without a row; anything with a path or a file extension is started exactly as written. Open URL: "https://www.example.com/one; https://www.example.com/two". Each opens in your default browser. Keep the space after the semicolon: a web address can have a semicolon of its own, and the space is how Lickety tells yours from one of those. Open folder: "C:\Reports; D:\Archive". Each needs its full path. Open file: "C:\Docs\Plan.docx; D:\Scans\Lease.pdf". Each needs its full path and its extension.

TARGET — where. A folder path, a file path, a program, or a web address. Windows variables work the way they do at a prompt: %windir%\system32\services.msc, %LOCALAPPDATA%\GreatScottLabs, %USERPROFILE%\Documents.

ON — untick to keep a row without it being matched.

RIGHT-CLICK A ROW to INSERT A ROW above it — your commands stay in the order you put them — or to copy it, paste a copy, or delete it. Delete on the keyboard asks first.

IMPORT… AND EXPORT… work as they do on Replacements, Vocabulary and QuickText: Export saves your commands to a file to keep or share; Import adds a file's commands to yours (a phrase you already have is skipped) or replaces yours with them, saving your current list first. A copy of someone's Commands.json imports too — it is the same format. Paths in a shared list are the other person's: check the targets after importing.

REFRESH START MENU re-reads the Start Menu and Windows' list of installed apps — do this after installing something so "open <that program>" works without a row. The button says Refreshing while it works, and the line under the list then shows how many apps were found and the time it finished.

APP LIST shows every name that "Open" will start, with a box to narrow it. It is a list to read, not to edit: to open something that is not on it, add a row with the program's path as the Target.

OPEN LOG opens commands.log in Notepad: one line per press, what was heard and what ran, kept for 90 days. You will not need it in normal use — the floating bar tells you when a press did not match — but it is there when a phrase keeps missing.

THE FLOATING BAR SAYS WHAT WAS HEARD. If your press matched nothing, it shows Heard: "..." — no command. Compare that with the Say column: the usual fix is adding the words the recogniser actually produced as an alternate, after a | in the same box. If the same mishearing turns up in your dictation too, a Replacement that corrects it also corrects the command: Replacements are tried on a press that matched nothing.

IT MATCHES ONLY WHEN IT IS SURE. A row matches on its whole phrase: one letter off is accepted ("open the dick" still opens the dock); two are not. Part of the phrase is not the phrase — "open the" matches nothing, and neither does "open the dock please" — and if what you said is one letter from two rows at once, nothing runs, on purpose: the wrong folder opening is worse than nothing opening. "Open <program>" is the one place the first words are enough: "open visual studio" starts Visual Studio 2022 when no other app's name begins that way.

"OPEN <PROGRAM>" FINDS NOTHING. The name has to match the name Windows shows for the app — "open Word" needs an app called Word. Click App list on the Commands screen to see every name Lickety knows, say that, or add a row with the program's path as the Target. Installed it just now? Click Refresh Start Menu.

IT MATCHED BUT DID NOT OPEN. The bar shows the command's name, but nothing appeared: the Target is wrong. A misspelled path, a web address without http:// in front of it, or a Windows variable that is not set on this machine. Open the row and check the Target exactly as Explorer or the browser would need it. An Open folder or Open file row whose target is not at that path opens nothing, and the bar says launch failed with the path it looked for — Explorer is never handed a path that does not exist, because its answer to one is to open some other folder.

THE HOTKEY DOES NOTHING AT ALL. Another program may own Ctrl+Numpad 5. Change it in Settings — the Voice command hotkey, under the paste-last and report-issue keys.

A note is a dictation you keep instead of sending somewhere else.

Open NOTES in the main window, press New note, put the cursor in the big box and press your dictation hotkey. Everything Lickety normally does to your words happens here too — your Vocabulary, your Replacements, QuickText, lists, numbers, dates, capitalization. It is the same dictation; it just lands in a note.

You do not have to give a note a title. If you leave the title empty, the list shows the first line of what you said. A title dictated into the Title box gets no full stop at the end — it is a title, not a sentence — and the same goes for the search box.

CREATE A NOTE BY VOICE — a voice note, if you like. Press the COMMAND hotkey (Ctrl+Numpad 5 unless you changed it) and say "create a note", "create note", "create a new note", "new note" or "make a note": Lickety comes up on Notes with a blank note open and the cursor in its title, from wherever you were. It is "open Notes" and the New note button in one phrase. If you are already on Notes with an unsaved note, it asks about that note first, as the button does. If the recogniser writes "node" for "note", that works too. The full list of what the command hotkey understands is under WHAT CAN I SAY? on the Commands screen.

SAVING. Press SAVE NOTE, next to the ← ALL NOTES arrow. If you leave a note with unsaved changes — Esc, the back arrow, or by opening another note — Lickety asks whether to save them, and CANCEL puts you back in the note. **A new note you have not typed anything into is thrown away without asking**, so changing your mind never leaves an empty note behind.

Leaving the Notes SCREEN from the left menu, or with Ctrl+Up / Ctrl+Down, asks the same question. Closing the window or Lickety saves what you have without asking, so nothing is lost that way.

WHERE YOUR NOTES LIVE. In the same file as your transcripts, on this PC, encrypted the same way. They are not sent anywhere and there is no account. If you have put a password on your history, a note is locked with everything else and shows as locked until you unlock it.

NOTES ARE NEVER TRIMMED. Transcripts roll off once you pass the history limit, because they are a record of text that went somewhere else. A note has nowhere else to be, so it stays until you delete it.

NOTES GO INTO EVERY BACKUP — AND THAT MAKES THEM DIFFERENT FROM TRANSCRIPTS. Your transcript history only goes into a backup if you tick the box for it. Notes always go in, even with that box off, because a note is something you meant to keep. They are written into the backup as READABLE text, so the backup can be restored on a different PC. If your backup folder is in OneDrive, Dropbox or iCloud, your notes go there too. **So a note is not the place for passwords or anything confidential.** A note that is locked (history locked with a password) is left out of a backup made while it is locked.

RESTORING A BACKUP KEEPS THEM. If you restore an older backup, any note you made after that backup was taken is kept rather than replaced, and Lickety tells you how many it kept. A note that was already in the backup comes back as the backup had it. A note in the backup that is missing here — deleted, or a new PC — comes back too, and Lickety tells you how many.

CATEGORIES, FOR WHEN THERE ARE TOO MANY TO SCROLL. Each note can carry one category, picked from a list you control — press EDIT LIST beside the picker to add, rename or remove names. It is a picker rather than a free-text box on purpose: three spellings of one category cannot be filtered, and filtering is what the field is for.

SORT AND FILTER THE BOARD. Above the cards, SORT orders them by most recent, title or category, and FILTER shows one category at a time. Each category in the filter carries the number of notes in it, so you can see where your notes actually are.

SEARCH THE BOARD. Type a word in the SEARCH box above the cards and press Enter or click SEARCH: the board shows only the notes that contain it, in their titles, their text or their categories, and the first one is highlighted. Click SEARCH again to move to the next one; the button counts them for you. Nothing is searched while you type. CLEAR puts every note back, and so does Esc in the box. A locked note is found by its title only; the words inside it are not read while it is locked.

THE BOARD FROM THE KEYBOARD. From the search box, TAB goes to SEARCH, TAB again to CLEAR, then to SORT, then to FILTER, and then to the first note; SHIFT+TAB goes back the same way. DOWN from the search box or from CLEAR goes straight to the first note. In SORT and FILTER the UP and DOWN arrows pick a value, and the board follows as you go. On the cards, the ARROW keys and TAB move from note to note and ENTER opens one. UP from a note in the top row, or SHIFT+TAB from the first note, goes back to the search box with its text selected, ready to type over.

SELECTING SEVERAL NOTES AT ONCE. A plain click on a card opens the note. CTRL+CLICK selects a card without opening it, and selects more as you keep Ctrl+clicking; SHIFT+CLICK selects every card from the last one you selected to this one — or, when nothing is selected yet, from the card with the keyboard focus (the first card, when you have just arrived on the board), so Shift+click on the last card selects them all; CTRL+A selects every card on the board (what the filter or search is showing). Selected cards are tinted. With cards selected, DELETE THIS NOTE under the board becomes DELETE 3 NOTES and asks once, listing them; the DELETE key on a card does the same, and so does RIGHT-CLICK → DELETE on any of the selected cards (the menu says DELETE 3 NOTES); COPY TO CLIPBOARD, or right-click → COPY, copies the selected notes one after another, a blank line between. Right-clicking a card that is NOT selected acts on that card alone. ESC clears the selection. Until you select something, those two buttons are greyed on the board, because there is no "this note" yet — inside a note they act on that note, as they always have.

A LOCKED NOTE CANNOT BE CHANGED OR DELETED. While your history is locked you can still see your notes on the board and sort them, because the category and the date are not part of the encrypted text — but DELETE, COPY, DUPLICATE and the category picker are all switched off, and EXPORT NOTES will tell you to unlock first. Deciding anything about a note you cannot read is guessing, and deleting one cannot be undone. Unlock your history and everything comes back.

RIGHT-CLICK A CARD ON THE BOARD for OPEN, COPY NOTE, DUPLICATE NOTE, the three EXPORT NOTE AS items, PRINT NOTE and DELETE. Duplicate makes an exact copy and opens it, so you can change the copy and leave the original alone. Copy Note puts the title AND the note on the clipboard. EXPORT NOTE AS MARKDOWN, EXPORT NOTE AS WORD DOCUMENT and EXPORT NOTE AS PDF each save that one note as a file you can read, named after the note's title. PRINT NOTE sends it to a printer you choose, laid out the same way: the title, then when it last changed and its category, then the note.

MERGE INTO ANOTHER NOTE moves the note you right-clicked to the bottom of a note you pick from a list, after one blank line — its title first, if it has one — and then deletes it. The note it went into keeps its own title and category. Lickety asks before it does this, and it cannot be undone.

KEYBOARD AND MOUSE IN A NOTE. TAB moves Title → Category → the note itself, and the category picker shows a coloured outline when TAB reaches it. A NEW note opens with the cursor in the title, so you can type a title, TAB to the category and TAB again to the note. An EXISTING note opens with the cursor in the note, because that is usually what you came back to change. ESC goes back to the board. RIGHT-CLICK in the title or the note for cut, copy, paste, delete and select all.

WEB LINKS IN A NOTE. Type or paste a web address that starts with http://, https:// or www. and it becomes a link. HOLD CTRL AND CLICK the address in the note to open it in your browser — the pointer turns into a hand while Ctrl is held. The row under the note also lists every link in it; a single click there opens one. A plain click in the note still just puts the cursor there, so you can edit the address. Only web pages open this way — never a file or a folder path.

KEYBOARD SHORTCUTS ON THE NOTES SCREEN. Hold ALT and the underlined letter on each button tells you its shortcut.

ALT+N New note ALT+E Export notes ALT+I Import notes ALT+A Back to All notes (inside a note) ALT+S Save the note you are in ALT+D Delete the note you are in ALT+C Copy the note you are in to the clipboard

A shortcut does nothing when its button is greyed out or not on screen — so ALT+S does nothing when there is nothing to save, and ALT+D does nothing on a locked note. Deleting still asks first.

IMPORTING. IMPORT NOTES adds notes from the CSV file that Export notes makes, or from Notes.json taken out of a Lickety backup — the way to bring your notes to a new computer, or to load a set someone shared. It only adds; nothing you have is changed or removed. A note with the same title and text as one you already have is skipped, and a note that was LOCKED when it was exported holds no text, so it is skipped and the message says to export it again with your history unlocked. Markdown, Word and PDF exports are for reading and are not imported. Restoring a backup brings your notes back too; Import is for taking only the notes.

EXPORTING. Press EXPORT NOTES to pick which notes go out. Everything is selected to begin with; UNSELECT ALL clears them, and FILTER BY CATEGORY narrows the list — tick as many categories as you like. If nothing is selected when you filter, the notes you filtered to are selected for you. The button says how many are going, so it always matches the file. Notes hidden by the filter are NOT exported, whether or not they are ticked.

SAVE AS picks the kind of file. CSV is the raw data, one row per note, for Excel: the title, the category, when the note last changed, and the note itself; a note with no title has an empty title column. MARKDOWN, WORD DOCUMENT and PDF are the same notes as something to read: each note's title as a heading, a line with the date and category under it, then the note, with a rule between notes and a heading for the whole set. A note with no title gets no made-up heading — its date line stands in. Files save to DOCUMENTS\LICKETY\EXPORTS, created for you the first time. A PDF is made by Windows' own MICROSOFT PRINT TO PDF printer, which asks where to save it.

Deleting a note asks first, and cannot be undone.

QuickText saves blocks of text you use over and over — your postal address, an email signature, a standard opening paragraph — and inserts one when you ask for it by name.

Say: INSERT QUICKTEXT OF MY ADDRESS. The words are replaced with whatever you saved under that name, across as many lines as you like.

Set them up under QuickText in the main window: give each one a short name and type the text into the box. Plain text only — bold and italics belong to whatever you paste into, not here.

EACH NAME CAN BE USED ONCE. Saying a name can only insert one block, so if you type a name that is already in the list, Lickety says so and clears the name on the new row. The text you typed stays — give it a different name to keep it.

IMPORT AND EXPORT work as they do on Vocabulary and Replacements. Export saves every block to a file you choose; Import asks whether to merge (names you already have are skipped) or replace, and a replace saves your current blocks to the ImportBackups folder first.

Say it however feels natural. "Insert QuickText of my address", "insert QuickText my address", "insert QuickText of the intro email" and "insert QuickText, intro email" all work — small words like "of" and "the" are optional, and a comma or full stop in between doesn't matter. So does using one mid-sentence: "Dear Bob, insert QuickText of my address, thanks."

The trigger phrase is two words on purpose. A single common word like "insert" would fire when you were only talking about inserting something. You can change it or add more under QuickText.

IF THE NAME ISN'T ONE OF YOURS, LICKETY SAYS SO. A message tells you which name it heard — "The QuickText 'test results' is not a known QuickText entry, so no QuickText was inserted." It never guesses at the closest name, because pasting the wrong block is worse than pasting none. If you said nothing but the trigger and the name, nothing is typed at all; if the rest of the dictation had other words in it, those are typed as you said them.

Punctuation looks after itself. If your text already ends with a full stop, you won't get a second one. If the QuickText is the last thing you said, nothing is added after it at all — it ends exactly the way you typed it. If it ends with a sign-off — "Thanks," on one line and your name on the next — nothing is added after the name, because a name isn't a sentence.

How this differs from Replacements: a replacement is a single-line correction applied silently to everything you dictate. QuickText is multi-line and only fires when you name it. Use replacements to fix what the recognizer gets wrong, and QuickText for text you'd otherwise retype.

Lickety keeps what you said. If a dictation looks like it may not have landed, a small panel appears near the clock showing the text with a COPY button — click Copy and paste it wherever you meant it to go.

It only appears when something looks wrong: the window changed while the text was being typed, or the text didn't turn up where it was expected. After a normal dictation you'll never see it.

It closes itself after a few seconds. The little circle beside the X is the countdown — hover the panel and it pauses, so it won't disappear while you're reading.

You can turn it off in Settings if you'd rather not have a panel to dismiss.

Either way, nothing is lost: every dictation is kept under Transcripts, and the paste-last shortcut re-inserts the most recent one.

Press ESCAPE while Lickety is listening and the recording is thrown away. Nothing is transcribed, nothing is typed, and nothing is added to your history or your statistics. The bar shows "Cancelled" for a moment and goes quiet.

This is different from stopping. Stopping — releasing the hotkey, tapping it again, or clicking the bar again — COMMITS: it transcribes what it heard and types it out. If you started by accident and someone was talking in the room, stopping would type that in. Escape is the way out.

Escape only does this while a dictation is listening or transcribing. At any other time it behaves normally.

One deliberate exception: if you were HOLDING your hotkey when you pressed Escape, Lickety keeps the Escape to itself. Otherwise your held modifiers plus Escape would form a shortcut you never meant to press, and the app you're dictating into could react to it. If nothing is held, Escape still reaches that app as usual — so cancelling while you're part-way through an Excel cell abandons the cell too, which is what most people expect Escape to do.

Escape works once it says "Transcribing…" too. If a transcription gets stuck, Escape abandons it and gives you the app back — and after a long wait the status bar says so itself.

Speech recognition is very good, not perfect. It helps to know WHICH kind of error you're looking at, because they have different causes and different answers.

1. UNKNOWN WORDS — names, products, jargon. Lickety has never heard of your customer, so it guesses something that sounds similar. This is the one you can genuinely fix: add the word to Vocabulary, and add a Replacement for the spelling it keeps producing.

2. RESHAPED PHRASES — the words are all real, the sentence reads fine, but it isn't what you said. "Calendar sync problem" becomes "calendar stink problem". These are the frustrating ones, because there's nothing obviously wrong on the page. They happen when the audio for a stretch is marginal and the recogniser fills the gap with something plausible.

3. SIMILAR-SOUNDING WORDS — "Anthropic" and "anaphylactic" can trade places, especially if both are in your Vocabulary. Seeding both tells Lickety both are likely, so it has to choose. If two of your terms sound alike and you rarely use one, leaving it out can improve the other.

4. DROPPED OR DOUBLED WORDS — usually a pause in the wrong place. See "Speaking for better accuracy".

WHAT ACTUALLY HELPS, in order: speak in complete phrases rather than fragments; add your real vocabulary; and read back anything that matters before you send it. That last one isn't a cop-out — it's the honest advice for every dictation tool, including the expensive cloud ones.

How you speak affects accuracy more than most people expect. A few things that make a real difference:

PHRASES, NOT WORDS. Speech recognition uses surrounding words to decide what it heard. A complete phrase gives it that context; isolated words don't. "Send the report to Fastenal" is far more reliable than "Fastenal" on its own.

MIND THE PAUSES. A pause mid-sentence can make Lickety think the sentence ended — you may see a short fragment split off as its own sentence, or a trailing word land oddly. If you need to think, it's better to stop the dictation and start again than to hold the hotkey through a long silence.

TRAILING SILENCE IS THE WORST CASE. Holding the hotkey for a second or two after you finish speaking is when the recogniser is most likely to invent something — repeating your last few words, or adding a phrase you never said. Release the hotkey promptly when you're done.

Lickety removes the clearest case of that: when the same phrase is repeated over and over at the end, the repeats are dropped and one is kept. It cannot tell where the invention started, only that a phrase repeated four or more times was not spoken that way — so releasing the hotkey promptly is still the fix, not this.

STEADY, NOT SLOW. Over-enunciating or speaking unnaturally slowly tends to make things worse, not better — it's further from the natural speech the recogniser was trained on. Speak as you would to a colleague.

TIREDNESS SHOWS. Later in the day, speech gets quieter, faster and less distinct, and accuracy drops with it. If results get noticeably worse over a long session, it's usually not the software changing.

MICROPHONE POSITION. A consistent distance matters more than an expensive microphone. Moving closer and further mid-sentence is worse than being consistently a bit far away.

"Wrong" has several different causes, and each has a different fix. Find the one that matches what you're seeing, then follow the matching link below:

• THE WHOLE SENTENCE IS NONSENSE — words you never said, invented words, random Capital Letters.

• A NAME OR TERM IS ALWAYS WRONG — your company, a customer, a coworker, an acronym.

• A WORD THAT SOUNDS THE SAME IS WRONG — read/red, sync/sink, write/right.

• THE RIGHT WORD, WRITTEN WRONG — correct word, wrong capitals or spacing.

• YOU DICTATED A SINGLE WORD and it came out as a different word.

• IT INVENTED SOMETHING you never said, out of nowhere, in an ordinary sentence. That happens.

The difference matters: a garbled sentence is an audio or pausing problem, while a single wrong word is usually just something not set up yet in Vocabulary or Replacements.

Every so often — even from clean audio, in a normal sentence, with everything set up correctly — the transcriber will produce something you simply didn't say.

A real example: "it also exits the app" came out as "it also, X is the tray house." Nothing was wrong with the microphone. Nothing was misconfigured. The sentence was ordinary and clearly spoken.

THIS IS NORMAL, AND IT ISN'T YOU. Speech recognition doesn't match sounds to words the way a dictionary does — it PREDICTS the most likely text for the audio it heard. Usually that's uncannily good. Occasionally it's confidently wrong, and it invents a phrase that sounds vaguely similar. Every tool built on this technology does it, including the expensive ones.

WHAT TO DO: re-dictate it. That's genuinely the answer — it's not a setting you're missing, and there's nothing to fix.

WHAT ACTUALLY HELPS: • Add the names it keeps missing to Vocabulary — that fixes the predictable failures (your customers, products, jargon). • Speak in full sentences and don't pause mid-sentence — those are the conditions where it invents most. • Read back anything that matters. That's true of any dictation tool.

What DOESN'T help: a bigger model (it can invent more confidently, not less), or hunting through settings. Random occasional nonsense is the price of the technology, not a fault in your setup.

Dictating ONE word on its own is unreliable — in Lickety and in every other dictation tool. It isn't a fault in the software; it's how speech recognition works.

WHY: a single word is a few hundred milliseconds of audio with nothing around it. Speech recognition is mostly CONTEXT recognition — the words before and after are what let it choose between things that sound alike. Say "sync" on its own and it's a coin flip between sync, sink, think, and thing. Even a person would ask you which one you meant.

FIX: speak a short sentence instead of a bare word — the extra words cost you nothing and give the recognizer something to work with.

But be realistic about what context buys you. Tested with Replacements emptied: "sync" in a full sentence still came out as "sink" some of the time. Context helps; it does not make homophones solvable, because the two words sound IDENTICAL. If a homophone matters to you, add a Replacement (tray → Replacements) — that's the tool for it, and it works every time.

Two related notes: • Adding the word to Vocabulary does NOT fix this. There's simply not enough audio to work with. (Tested with the vocabulary emptied: same result.) • A bigger, slower model doesn't fix it either — it can make it worse, because a larger model leans harder on context that isn't there.

If you need a one-word command often, dictate it as a short phrase and let a Replacement shorten it (tray → Replacements).

Windows 11 replaced the thirty-year-old Notepad with a rewritten app (it's now built on the same text engine as Word), and it handles synthetic input differently from other Windows apps. Lickety accounts for this, and your dictation lands correctly — including lists that need separate lines.

THE MENU BAR FLASHES WHEN YOU PRESS THE HOTKEY. If your hotkey contains Alt — Ctrl+Alt is the default — Notepad's File/Edit/View menu bar lights up for a moment when you press it. That's Windows, not Lickety: in any app with a menu bar, Alt on its own is the keyboard shortcut that ARMS the menu, and Windows sees the Alt in your chord before it sees the rest of it. Notepad is simply the app you are most likely to notice it in.

IT IS COSMETIC. The menu closes on its own, your text still goes in, and nothing is lost — it is a flicker, not a failure.

TO STOP IT COMPLETELY, choose a hotkey with no Alt in it (Settings → Hotkey). Any Alt-free chord removes the flash entirely.

BUT READ THIS FIRST IF YOU USE REMOTE DESKTOP. Over a full-screen Remote Desktop session most key combinations are captured by the remote machine before Lickety can see them, and Ctrl+Alt is the most reliable chord there is. If you dictate into Remote Desktop, keeping Ctrl+Alt and living with the flash is the better trade — a hotkey that doesn't fire is a much bigger problem than a menu that blinks.

IF TEXT ITSELF LANDS WRONG: the same dictation into Word, Outlook, a browser, Notepad++, or VS Code works — none of them have Notepad's quirks. And Ctrl+NumPad1 (paste last transcript) re-inserts the kept, correct text.

A SHORT DICTATION MAY NOT GET ITS CAPITAL BACK. Lickety decides whether you are starting a new sentence or adding to one by reading the couple of characters just before the cursor. Notepad does not reliably hand those over, so in Notepad the answer is often "I cannot tell".

You notice it with Lowercase short dictations turned on. Finish a sentence, stop, then dictate two or three words: in Word the new sentence gets its capital, and in Notepad it usually stays lower case. Nothing is lost and no words change — only the first letter — and typing over it is one keystroke.

This is the same read behind two other behaviours, so they come and go together in Notepad: whether a space is added after a full stop, and whether a continued sentence keeps its lower case. **It is a limit of what the app tells us, not a setting you can change** — the alternative is guessing, and a guess would put a capital in the middle of your sentences.

Anything else odd in Notepad is worth reporting (Home → Report an issue) — the report carries the technical details automatically, and that's what makes it fixable.

If a whole sentence comes out as nonsense — invented words, things you never said, or oddly Title Cased Words — the transcriber lost the thread. Three common causes, in the order worth checking:

1. YOU PAUSED MID-SENTENCE. The transcriber works on the whole clip at once and uses the earlier audio to predict the rest. A long silence in the middle leaves it with nothing to go on, and it starts inventing — often Title Casing as it drifts. Example: pausing after "version sixty fixed" produced "...on a Couple of Testment of Duns of Bar." FIX: don't pause mid-dictation. Release the hotkey, then press it again and continue. Lickety is built for this — it will join your sentence correctly without adding a stray period or capital.

2. THE WRONG MICROPHONE. Often a webcam or far-field mic instead of your real one. Windows can switch your default mic silently when a device connects, so this can start with no change on your part. FIX: Audio → Input device → pick your actual microphone. Then check Windows Sound settings → Input: speak and watch the level meter. If it barely moves, the mic is too quiet or too far away.

3. HOW YOU'RE SPEAKING — the cause people skip past. Mumbling, running words together, talking fast, turning away from the mic mid-thought, background noise, or letting the last few words of a sentence trail off. The transcriber only has the sound that reached the microphone, and it can only be as clear as that was — it isn't reading your mind. FIX: speak at a normal, even pace, right through to the end of the sentence. You don't need to over-enunciate or slow down. Most bad accuracy turns out to be sentence endings going quiet. Clean input matters more than any setting in this app. A note on models: switching to a bigger, slower model does NOT reliably fix this. A larger model leans harder on context to fill gaps, so on poor or interrupted audio it can invent MORE confidently, not less. Fix the audio and the pausing first — that's where the gains actually are.

If a specific name comes out wrong EVERY time — your company, a customer, a colleague, an acronym — the transcriber has simply never heard that word. It isn't misunderstanding you; it's guessing at a sound it doesn't know, so it reaches for ordinary words instead (a real example: "Fastenal" became "And all for you and us").

You cannot fix this with Replacements — the transcriber invents a DIFFERENT wrong guess each time, so there's nothing consistent to replace.

FIX: add the word to Vocabulary (tray → Vocabulary). That tells the transcriber the word exists BEFORE it listens, so it can recognize it rather than guess.

Tips: • Add the names you actually use — customers, vendors, products, coworkers, internal jargon. Don't pad it with words the recognizer already gets right; a focused list works better than a long one. • There IS a size limit: the list is passed to the transcriber as a prompt with a ceiling of about 224 tokens, and terms past that aren't seeded. The Vocabulary panel shows a CAPACITY METER so you can see how close you are — it stays muted with room to spare, and turns amber then red as you approach and pass the limit. Most real lists sit well under it. • Tokens aren't words. A common word is often 1 token; an unusual name gets split into pieces. Reckon on 2-3 tokens per term as a rough average — very roughly 60-80 terms — with long, unusual names costing more. The meter estimates this for you, so you don't have to count. • If the meter goes red, trim terms the recognizer already handles without help; keep the ones it genuinely needs taught. • Write a term the way you want it to appear; that nudges the formatting too. • Very short dictations (one or two words) give the transcriber little to go on, so it may lean on your vocabulary too hard. Speak a full sentence when you can. • If two of your vocabulary terms are very similar (e.g. two names that start the same way), the recognizer can occasionally BLEND them — say one and get a mix of both. It's uncommon, and it's fixable the same way as any stable slip: add a Replacement mapping the blended output to the word you meant. This isn't the list being too full (see the capacity meter) — it's just two look-alike terms, so one disambiguating Replacement sorts it.

Some words sound IDENTICAL and no transcriber can tell them apart from audio alone — "marked as read" vs. the color "red", or "sync" vs. "sink". This is different from an unknown name: the word came out as a real, correctly-spelled word — just the wrong one.

FIX: use Replacements (tray → Replacements).

• If you never mean the other word, use a plain replacement: sink → sync.
• If you use BOTH words, fill in 'Only after' so the rule fires only in context: Heard=red, Replace with=read, Only after=unread to. That fixes "unread to red" and never touches the color "red" elsewhere.

Some homophones aren't worth a rule — their/there/they're, your/you're, its/it's depend on grammar that changes every sentence, so a rule would be wrong as often as right. Those are easier to catch on a read-through.

At the bottom of the Replacements screen is a short second table: "Corrections that ship with Lickety". These are ours, not yours.

They exist for the handful of things the recognizer gets wrong for everybody, not just for you — for example, "diagnostics" coming back as "diamostics".

THREE THINGS WORTH KNOWING:

• They are never written into your own Replacements file. Your list stays yours; export it and you will not find them in it.
• They run BEFORE your rules, so your own rules see the repaired text and still win.
• Untick one to switch it off. It stops on your very next dictation, and you can tick it back on whenever you like — nothing is deleted.

If you write a rule of your own for the same phrase, ours steps aside automatically and the table says so.

The table is deliberately short. A correction only ships once the mistake has actually been seen — guessing at them is how an app starts rewriting words nobody misheard.

WHAT WILL NEVER SHIP HERE: sound-alikes between a letter or a digit and a word — "C" for "see", "4" for "for". Nothing in the text can tell them apart. "Item 4 payment terms" and "item for payment terms" are both ordinary English, and a rule that picks one is guessing on your behalf, so Lickety does not make it. If you hit one of these often, a bigger recognizer model (Settings → Model) is the lever, not a correction — and for a phrase only YOU say, write your own Replacement with an "Only after" context.

There is no "delete everything" button, on purpose.

Transcripts are removed by exactly one thing: your retention setting, which trims the oldest once you pass the limit you chose. Nothing else deletes them, and nothing in Lickety deletes them all at once.

The Clear button beside the search box clears the SEARCH and puts the full list back. Your transcripts are not affected by it.

If you want everything gone, that is an uninstall question rather than a button question — a single click that destroys months of history, one keypress away from the search box, is a worse risk than the problem it solves. Uninstalling Lickety (Windows Settings → Apps) asks whether to also remove your data — settings, lists, Notes and history — and the answer is No unless you choose Yes.

To delete individual entries, use the list itself. To keep a copy of everything, use Backup with "Also include my transcript history" ticked.

If the transcriber HEARD the word correctly but wrote it the wrong way — "do it best" instead of "Do It Best" — that's a formatting fix, not a recognition problem.

FIX: add a Replacement (tray → Replacements): Heard = do it best, Replace with = Do It Best.

TWO THINGS ABOUT CASE, because they are not the same:

• The HEARD side ignores case completely. "do it best", "Do It Best" and "DO IT BEST" are the same rule — there is no point retyping your rules in lower case.
• The REPLACE WITH side is used exactly as you type it, with one exception: if you write it in lower case and it lands where a sentence capital belongs, Lickety supplies the capital. So write it lower case and let the app decide, or write the capital yourself and it is left alone. Mid-sentence, your lower case is what you get.

A SHORT ALL-CAPS TRIGGER ALSO MATCHES ITS SPELLED-OUT FORM. If your Heard side is something like TWP or KIB — no spaces, two to six characters, no lower case — then TWP, T-W-P and T.W.P. are all the same rule. You do not need one row per spelling, which matters because the transcriber picks the spelling, not you: the same word said a little slower comes out as separate letters. Spaces are NOT treated this way on purpose, or a rule on A1 would fire every time you said "a 1".

A COMMA BETWEEN THE WORDS DOES NOT MATTER. The transcriber writes a comma where it heard a pause, and it does not do so the same way twice: "pound NA" one take, "pound, NA" the next. A rule on either spelling matches both, so you need one row, not two — and the same goes for the phrases Lickety listens for itself ("compose email", "line break", the list phrases, the time phrases). A comma at the very end of a Heard phrase, or a full stop anywhere, is still taken as typed.

Vocabulary and Replacements do different jobs, and the hardest names need BOTH — they work as two stages: • VOCABULARY gets the transcriber to HEAR the word (before it listens). • REPLACEMENTS fix how it's WRITTEN (after it listens).

A real example of the two stages together: "Fastenal" with nothing set up came out as wild, different guesses each time (unfixable by a replacement, because there's no consistent wrong answer to catch). Adding it to Vocabulary got the recognizer into the ballpark — now it reliably produced "Fastenl", a small, STABLE miss. A single Replacement (Fastenl → Fastenal) then closed the gap. Vocabulary turned a wild, unpredictable miss into a small predictable one; the Replacement fixed the predictable one. That's the pattern for any stubborn name: teach it in Vocabulary first, then add a Replacement only for whatever small, consistent slip is left.

Lickety inserts text two ways:

• PASTE (default): fast, works in most apps (Word, Excel, Outlook, Notepad, browsers).
• TYPE KEYSTROKES: for apps that reject paste — terminals, remote desktop sessions, some older or secure fields, and Python windows.

IF ONE APP WON'T TAKE TEXT, FIX IT FOR THAT APP — not globally. Go to Settings → Per-app insertion method and add a row: theapp.exe → keystrokes. This changes nothing anywhere else.

To find the process name: Task Manager → Details, and look for the app in the Name column (e.g. mstsc.exe, WINWORD.EXE).

Why per-app rather than the global setting: paste is faster and works in most places, so switching everything to keystrokes to fix one app trades a small problem for a broad one. Change the global default (Settings → Text insertion) only if MOST of the apps you dictate into reject paste.

Already handled for you: • mstsc.exe (Remote Desktop) — ships set to keystrokes. • Python apps (python.exe / pythonw.exe) — auto-detected. Electron-based apps (some chat and editor apps) can also be occasionally unreliable with paste; if you see text go missing in one, add it to the per-app list.

Lickety dictates into a Remote Desktop session with NOTHING INSTALLED ON THE REMOTE MACHINE. Run it locally and it types into the session as if you were typing yourself — which also means nothing to keep in sync, and it works where you aren't allowed to install software.

This works out of the box: mstsc.exe (Remote Desktop) ships in Settings → Per-app insertion method, set to keystrokes. Keystrokes are Remote Desktop's native channel; pasting goes through a separate clipboard-redirection layer that isn't reliable.

Notes: • Any hotkey works, full screen or windowed, except one with the Windows key. When the RDP window is FULL SCREEN it watches the keyboard itself; since version 589 Lickety puts its own keyboard watcher back in front of it while the session is in front, so allow about half a second after going full screen before the first hotkey. • No microphone redirection needed. Your local mic is doing the work. • Other remote-access clients (Citrix, VMware Horizon, VNC) have not been tested. If yours doesn't accept text, add it under Settings → Per-app insertion method with keystrokes and it will likely work the same way.

If you'd rather run Lickety ON the remote machine instead, that also works: redirect your mic (RDP client → Show Options → Local Resources → Remote audio → 'Record from this computer').

Your hotkey: anything but the Windows key works. A chord with a letter or number key (Ctrl+NumPad1) works in a full-screen session too since version 589, about half a second after the session takes the screen. Windows-key chords don't work over RDP at all, and Lickety won't let you set one.

If no hotkey works at all in your setup, you don't need one: turn on the always-visible status bar and click it to start and stop. That path doesn't involve the keyboard, so nothing can intercept it.

Lickety's window can be driven from the keyboard:

• CTRL+DOWN / CTRL+UP — move to the next / previous item in the left menu (Home, Transcripts, Replacements, and so on), from anywhere in the window. Change screens without reaching for the mouse.
• EXCEPT IN A TEXT BOX. While the cursor is in a box you type in, the Ctrl+arrow keys do what they do in any editor — move a word or a paragraph at a time. Press Tab or click outside the box first to change screens.
• In any of the editable lists (Replacements, Vocabulary, Auto-Lowercase, Lists): HOME jumps to the top of the list, END jumps to the bottom (where new entries are added). CTRL+HOME / CTRL+END do the same from anywhere. These work as soon as the list opens.
• While you're editing a cell, Home and End work on the TEXT you're typing as normal; the list-jump only applies when you're not mid-edit.
• Click a column header to sort a list; ENTER in a list commits the row and starts a new one; right-click a row to Insert a row above it, Copy, Paste as a new row, or Delete it.

• Edits take effect on your next dictation as soon as you finish a row — you don't have to click Save first (Save is still there, and still shows a "Saved" confirmation).

Your dictation hotkey is separate and set in Settings.

FROM OUTSIDE THE WINDOW. Three hotkeys open the three Tools screens from any program while Lickety is running: CTRL+SHIFT+C for Commands, CTRL+SHIFT+N for Notes and CTRL+SHIFT+T for Transcribe, unless you changed them in Settings. The window comes to the front — or opens — on that screen. Lickety does not take these keys away from the program you are in, so in Edge or Chrome Ctrl+Shift+N also opens a private window, Ctrl+Shift+T also reopens a tab and Ctrl+Shift+C also opens Inspect, and in File Explorer Ctrl+Shift+N also makes a new folder. Pick other keys in Settings if that gets in the way, or Clear the ones you don't use.

The default is CTRL+ALT, and for most people there's no reason to change it.

If you do change it, two things are worth knowing:

• THE WINDOWS KEY CAN'T BE USED. Windows-key chords never reach a full-screen Remote Desktop session, and they fail with no error at all — the hotkey simply does nothing, which is impossible to diagnose. Lickety won't let you set one.

• IF YOU DICTATE OVER REMOTE DESKTOP, any chord without the Windows key works, full screen included (version 589 on). In a full-screen session allow about half a second after the session takes the screen before the first hotkey — see the Remote Desktop topic for why.

• TWO HOTKEYS CAN'T SHARE THE SAME KEYS. If the keys you press are already the dictation hotkey or one of the others (Paste last transcript, Report an issue, Voice command, Open Commands, Open Notes, Open Transcribe — and Great transcription, which is on the screen only while the diagnostic controls are shown), Lickety says which and keeps the old one. Adding a key to another hotkey's keys is allowed.

• ALTGR IS NOT CTRL+ALT. On a keyboard layout with an AltGr key (UK, German, French and others), Windows reports AltGr as Ctrl plus Alt. Lickety ignores that Ctrl, so typing € or an accented letter never starts a dictation. Ctrl+Alt still works with either Alt key, as long as you press Ctrl.

A note on Ctrl+Shift: it works everywhere, but because it's only modifiers, it also fires on shortcuts that begin with it — Ctrl+Shift+T, Ctrl+Shift+N, Ctrl+Shift+Esc. Those will start a dictation as well as doing their usual job. Ctrl+Alt collides far less, which is why it's the default.

A note on ALT: any chord containing Alt makes the menu bar flash in apps that have one — you'll see it in Notepad. Windows treats Alt on its own as the key that arms the menu, and it sees that Alt before it sees the rest of your chord. It's cosmetic, the menu closes itself, and your text still goes in. An Alt-free chord removes it — but if you dictate over Remote Desktop, keep Ctrl+Alt anyway: a hotkey that doesn't fire is a far bigger problem than a menu that blinks.

The hold-vs-tap timing threshold is adjustable in Settings if a hold is being read as a tap or vice versa.

If no hotkey suits you at all, you don't need one — turn on the always-visible status bar and click it to start and stop.

Transcription runs entirely on your computer (no cloud), so speed depends on your CPU. The default model, base.en, is the fast choice (~1.5s for a short sentence). small.en is more accurate but slower.

Switch models in Settings → Model. On a typical CPU, base.en is the best balance; choose small.en only if you need maximum accuracy and accept the wait. Model size is the main speed lever — the rest is fixed by your CPU.

If the transcriber consistently mis-handles a name, company, or term — e.g. lowercasing 'do it best', or writing a heard word the wrong way — add a Replacement.

Open it from the tray menu → Replacements (its own screen). Enter what's heard (e.g. 'do it best') and what it should become ('Do It Best'). The replacement is inserted exactly as you type it, so you control capitalization (including ALL CAPS), and it matches the whole phrase so it won't affect unrelated words.

A RULE MATCHES INSIDE A SERVER NAME, A WEB ADDRESS OR A PATH. Dots, slashes and hyphens count as word boundaries, so Heard "akmeparts", Replace with "acmeparts" fixes UPS.Akmeparts.com, mail.akmeparts.com and akmeparts.com alike — one rule for the whole domain, and the capital letter it was heard with is kept — and it covers jane@akmeparts.com and https://akmeparts.com too, so one bare rule replaces an @ rule and a dotted one. THE RULE OF THUMB: start the rule at the first LETTER of the misheard word, never at the punctuation around it. ".akmeparts.com." with the dots can never fire, because a rule that starts with a dot demands a space before it and there is a letter there instead.

For a name the recognizer gets wrong in a DIFFERENT way each time, a Replacement alone can't help — add it to Vocabulary first, then use a Replacement for any small consistent slip that's left.

FINDING ONE IN A LONG LIST. Two controls at the top, and they do different jobs. SHOW narrows by category. SEARCH shows only rules containing what you type, in any column — type "ac" and you get Acme along with anything else containing those letters, wherever they sit in the word. They combine, so you can search inside one category. Escape clears the search box, and CLEAR resets both at once.

THE CATEGORY COUNTS FOLLOW THE SEARCH. With a search active, Show lists only the categories your results are actually in, with the counts from those results — because a number beside a category is a promise about what picking it will show you.

SETTING A CATEGORY FROM THE KEYBOARD. Tab to the Category cell, press Down or Alt+Down to open the list, arrow to the one you want, then Enter to choose it. That first Enter belongs to the dropdown; press Enter again to finish the row and start a new one.

EACH RULE CAN BE IN THE LIST ONCE. If you type a Heard that is already there — with the same Only after — Lickety tells you what the existing rule changes it to and removes the new row, because only one of the two could ever be used. If you edited an existing rule into a copy of another, your edit is undone instead. Capitals and extra spaces don't make it different: "akme" and "AKME" are the same rule. The same Heard with a different Only after is a different rule and is allowed. Vocabulary terms and QuickText names are checked the same way.

Everything you teach Lickety — vocabulary, replacements, lists, QuickText and every setting — lives in a hidden application-data folder. That's a folder most backup software never touches, so Lickety keeps a copy somewhere yours already looks.

A backup is EVERY FILE IN THAT FOLDER, not a list of files Lickety knows the names of. Anything you have put there yourself goes in too, and so does anything a future version adds. The only things held back are the ones that hold your words or your voice, and those are the transcript setting below.

SUB-FOLDERS ARE NOT INCLUDED. The folder Lickety writes its diagnostic recordings into is one of these, and it can run to hundreds of megabytes that change every few minutes — backing it up made every archive enormous and slow for no benefit, since those recordings are working files. If you keep something of your own in a sub-folder there, copy it up into the main folder or back it up yourself.

One snapshot a day, kept for ten days, in Documents\Lickety\Backups. The day's file is updated as you work, so it stays current without filling the folder with near-identical copies. Ten days means you can go back to before a mistake, not just before the last one.

IF YOU SEE TWO FILES FOR ONE DAY, one of them holds something the other does not. Lickety only removes an earlier archive when the new one contains everything that was in it — so if you turn the transcript setting below OFF, the archive that still has your history is kept rather than replaced by one without it. That is the only reason a day ever has more than one file.

BACK UP AUTOMATICALLY — on by default. It runs on its own; turn it off and nothing is written unless you press the button. The Backup screen shows when the last one was written and where, and there's a Back up now button for the moment before you try something risky.

TO PUT SOMETHING BACK, use Restore. Your current files are saved first, so restoring the wrong backup can itself be undone — the copy is named before-restore- and the date. THAT COPY ALWAYS CONTAINS YOUR TRANSCRIPT HISTORY, whatever the setting below says, because it never leaves this PC and it is the only thing standing between you and a restore you did not mean. Lickety closes afterwards; reopen it to load what you restored.

RESTORE READS THE BACKUP BEFORE IT ASKS YOU ANYTHING. It tells you how many dictations that backup holds and how many are on this PC, so "an older backup" stops being a guess. If restoring would remove dictations that exist only here, it says how many, it ticks KEEP THE DICTATIONS ALREADY ON THIS PC for you, and it makes you type a word before it will replace them. If there is nothing to lose — a new PC, or a rebuild — it just restores.

YOU CAN RESTORE EVERYTHING EXCEPT YOUR HISTORY. That is what the tick box does, and it is the one to use when you have wrecked a list or a setting and want it back without giving up the dictations you have made since.

IF YOUR HISTORY IS LOCKED, UNLOCK IT FIRST. A restore that replaces your history has to carry the notes on this PC across, and it can only read them while history is unlocked, so it stops and asks. Ticking the box to keep the dictations already on this PC also keeps your notes, and needs no unlock.

IF A RESTORE STOPS PART-WAY, it says how many files it had already put back and where the before-restore copy is, then closes. Restoring that copy puts everything back as it was.

IMPORT, THEN REPLACE, ON A LIST SCREEN SAVES YOUR LIST FIRST. The list as it was goes to the ImportBackups folder beside your settings, and the message after the import names the file. To undo, Import that file and choose No (replace).

YOUR NOTES ARE ALWAYS INCLUDED, as readable text — see Notes. That is why a note is not the place for anything confidential.

YOUR TRANSCRIPT HISTORY IS NOT INCLUDED unless you ask for it. It holds every word you have ever dictated. Lickety never sends it anywhere, but if your backup folder sits inside OneDrive, Dropbox or iCloud, including it means a copy gets uploaded by THAT service. Worth being a decision you made rather than one that happened.

THE SAME SWITCH COVERS EVERY OTHER FILE WITH YOUR WORDS IN IT: the voice-command log (what you said after the command hotkey), the About-screen report that quotes your sentences (phrase usage in history), the diagnostic log if you turned it on, and any saved recordings. The everyday log, dictation.log, holds timings and counts only, never your words, so it is always included.

IT IS ALSO ALMOST ALL OF THE SIZE. Everything else you have taught Lickety — your vocabulary, replacements, QuickText, every setting — is a few megabytes. The history can be hundreds. So an archive that suddenly looks far smaller has almost certainly just had the history left out, not the rest of your folder.

You can change the folder to anywhere your own backups reach.

Everything you dictate is kept in one file on this PC. Nothing is ever uploaded. This screen is about who could read that file if they got hold of it.

ENCRYPT MY HISTORY — on by default — and ALSO REQUIRE A PASSWORD TO READ IT — off by default — are the two switches on this screen. The first is the protection you already have; the second adds the password described below.

IT IS ENCRYPTED, AND YOU DO NOT HAVE TO DO ANYTHING. Windows holds the key and ties it to your account, so Lickety opens your history normally and you will never be asked for anything. Somebody who copies the file off this machine, or signs in as a different user on it, gets nothing they can read.

What that does not cover: software running as you, while you are signed in. No encryption of a file at rest covers that, ours included. It is worth saying plainly rather than leaving you to find out.

A PASSWORD IS OPTIONAL AND OFF UNLESS YOU TURN IT ON. Most people do not need it, and it gives you a way to lose your history that you do not have today. It is there for work you may have to account for — patient notes, client matters — where being able to say nothing on the machine opens your history without you is the point.

With a password set, your history locks after a period of inactivity you choose. DICTATION CARRIES ON WORKING WHILE IT IS LOCKED. Everything you say is still recorded; Lickety simply cannot read any of it back until you unlock, and then it is all there. An empty Transcripts screen while locked is not lost history.

YOUR NOTES GO READ-ONLY WHILE IT IS LOCKED. They stay on the board and you can still sort them, but you cannot change, copy, duplicate, delete or export one until you unlock — deciding anything about a note you cannot read is guessing, and deleting one cannot be undone.

TYPING THE RECOVERY CODE BACK IN: the dashes are optional and capitals do not matter. It is stored as letters and digits only, so however you type it is fine. Tick "Show what I'm typing" on the unlock screen if you want to see it go in — a 25-character code copied off a piece of paper is the worst thing to type blind.

CHANGING YOUR PASSWORD ISSUES A NEW RECOVERY CODE AND RETIRES THE OLD ONE. This is the usual reason a code is refused. An old code fails in exactly the same way as a wrong one, so the unlock screen tells you the date the current code was created — if the one you are holding is older than that, it is not the one. Turning the password off and on again counts as a change, and so does every password change after it.

THE RECOVERY CODE DOES EVERYTHING THE PASSWORD DOES. It opens your history, it can set a new password, and it can turn the password off — because a code that only unlocked would leave you entering 25 characters for ever, and "the way back in if you lose your password" is what it is for. Using it to set a new password issues a new code and retires the one you used, which is the right way round: the moment you recover, the code you had written down stops working.

YOUR RECOVERY CODE IS THE ONLY OTHER WAY IN. It is shown once, when you set the password. Put it in your password manager or write it down and keep it somewhere other than this PC — a code saved only on the machine it unlocks is no help when that machine is what died. If you lose both the password and the code, your history cannot be recovered, by us or by anyone. That is what makes it worth having.

THE UNENCRYPTED COPY IS REMOVED ONCE ENCRYPTION IS CONFIRMED. When your history is first encrypted, Lickety keeps a plain copy of it for a moment, in case the conversion fails. As soon as every record checks out as encrypted, that copy is deleted, so it never sits beside your encrypted history or goes into a backup. If the check cannot be completed, the copy stays until it can.

BACKUPS. Without a password, your backup is written so it can be read — it has to be, or a backup made on this PC would not open on your next one. You chose where it goes; treat it like any other file holding your work. With a password, the backup stays encrypted and opens with the same password or recovery code on any machine.

If a dictation comes out wrong, you can send a report with the details attached.

For the one that just happened, use your report-an-issue shortcut.

FOR AN EARLIER ONE, open Transcripts, right-click the line you want and choose "Report an issue with this one". That matters because problems are often only obvious after you've moved on — by then the shortcut would report the wrong dictation.

THE NOTE IS REQUIRED. Say what went wrong in the "What went wrong?" box — even one line. If you press Report with the box empty, the window stays open and says so; no report is written and the dictation is not counted as a problem until you do.

A report includes your note, what was heard, what was inserted, which application it went to, and how the text was delivered. It never includes your audio unless you have turned on diagnostic capture, and nothing is sent anywhere automatically.

TO SEND ONE TO US, save it as a file when the report window offers to, and attach that file to an email to support@greatscottlabs.co. The file's name starts with Lickety, which is how we know which program it is about; please leave the name as it is.

THE ADDRESS IS ON THE ABOUT SCREEN three ways: click it and a new message opens in Windows' default mail program with the version in the subject line; press COPY beside it to put it on the clipboard for the mail program you actually use; or read it off the screen. If Windows has no mail program set up for links, the click copies the address instead and says so. The Great Scott Labs logo at the top of the same screen opens the page listing our other tools, finished and coming, in your browser.

If Lickety itself misbehaves — doesn't appear when Windows starts, or closes unexpectedly — two files in your Lickety folder help: startup.log records every launch with how long the computer had been running, and crash.log records anything that went badly wrong.

Stats (tray menu) shows words dictated today / this week / month / all-time, total dictations, your average speaking speed, and a chart of words per day (the last 14 days, the last 30, or month by month). All of it is stored only on your computer — nothing leaves the machine. History begins from when the stats feature was added.

Everything runs locally. Speech is transcribed on your own machine by Whisper — no cloud, no account, no internet needed, no subscription. Dictated text is never sent anywhere. Stats and settings are stored only on your computer. After dictating, your text is left on the clipboard, unless you turn on RESTORE MY CLIPBOARD AFTER DICTATING on Home, which puts back what was there. Either way, what Lickety pastes is kept out of Windows clipboard history (Win+V) and is never uploaded to the cloud clipboard.

Some words defeat even the vocabulary feature — usually long words made of smaller real words (the recognizer confidently hears the common words instead), or unusual character, medical, or foreign names. The tell: you say it three times and get three DIFFERENT wrong answers, and vocabulary doesn't help.

The fix — a shorthand alias: pick a short, easy word the recognizer always gets right, and add a Replacement mapping it to the hard word. Example: say "snuffy" with a replacement "snuffy" -> "Snuffleupagus", and you get the full word perfectly, every time. You've traded an impossible word for an easy one.

Pick an alias that's short, easy to say, and something you'd never dictate on its own (so it never fires by accident). This works great for character names, complex medical terms, and any word that's important to you but hopeless for the recognizer.

WHERE IT IS. Your vocabulary is vocabulary.json beside your settings file — its own file, so a backup from another PC can bring the words across without bringing that PC's hotkeys and window position too. It goes into every backup; restoring an older backup adds the words it holds and never removes yours.

IF YOUR REPLACEMENT IS A SHORT ABBREVIATION THAT EXPANDS INTO A SENTENCE, put it in the "Shorthand" category. This is the single most useful thing to know about Replacements and it is easy to miss, because a replacement with no category still works — it just loses one piece of cleanup.

WHAT THE CATEGORY DOES. When a shorthand expands into a whole sentence, the abbreviation you spoke may arrive from the recognizer with a comma after it — "TWP," one time and "TWP" the next, from the same spoken word. Lickety cannot control which one it gets. So for replacements in the Shorthand category, when the replacement text is a sentence, a trailing comma left behind after the swap is turned into a period. Without the category, you get a sentence that ends in a comma about half the time.

IT IS DELIBERATELY LIMITED TO THAT CATEGORY. The same cleanup applied to every replacement would damage the ones that are meant to end in a comma, and it would reach things that only look like sentences — a phone number, an email address. Narrowing it to a category you opt into is what makes it safe.

PICK AN ALIAS YOU WOULD NEVER SAY ON ITS OWN. Short letter-and-digit forms like "A2" also match the same letters and digits separated by a dot or a dash, so "A2", "A.2" and "A-2" all fire the same rule. That is what makes a spoken acronym work at all, and it is why the alias should be something you have no other use for. Two to six characters is the range that behaves this way.

IN SHORT — abbreviation that expands to a sentence: Shorthand category. Ordinary word-for-word correction like "granger" to "Grainger": no category needed.

A SPOKEN NAME LIKE tbl_selected_order OFTEN ARRIVES AS TBL_selected_order: the recognizer writes a short prefix in capitals because it looks like an acronym. A replacement matches whole words, and to a replacement the whole of TBL_selected_order is one word — so a rule on "TBL" would normally not fire inside it.

TWO CATEGORIES ARE ALLOWED TO FIRE AT THE START OF SUCH A NAME. A rule in the Capitalization category (Heard "TBL", Replace with "tbl") or in the Shorthand category fires when its letters begin a word and stop at an underscore. So "TBL" → "tbl" in Capitalization turns TBL_selected_order into tbl_selected_order, and leaves TBLX_thing and sub_TBL_x alone — the letters must start the word and must stop. One rule per prefix: tbl, qry, rpt, mcr, or whatever your naming uses.

WHY ONLY THOSE TWO. A Capitalization rule can only change the case of the letters it matched, and a Shorthand rule is one you typed to mean "expand this" — neither can rewrite the rest of the name. An ordinary correction such as "sink" → "sync" is kept out of identifiers on purpose, so a field called sink_rate is never quietly edited.

AT THE START OF A SENTENCE the first letter is capitalized as for any sentence; mid-sentence, which is where a name like this is nearly always typed, it comes out exactly as the rule says.

THE ADDRESS IS ON THE ABOUT SCREEN: support@greatscottlabs.co. Click it to open a new message in your mail program with the version already in the subject line, or press Copy beside it and paste the address wherever you write mail.

WHAT TO SEND. Say what you did, what you expected, and what happened instead. If a dictation came out wrong, use Report an issue on the Transcripts screen for that entry first — it writes a small file with the exact text the recognizer heard and the settings in force, and that file answers most questions in one go. Attach it to the mail.

WHAT NEVER LEAVES YOUR PC ON ITS OWN. Lickety sends nothing anywhere by itself. The report file is written to your disk and goes only where you send it.

Right-click the tray icon → About to see the version number. The version also shows in the tray icon's tooltip. Use this to confirm a new build is actually the one running.

Four settings live on the Home screen.

APPEARANCE — Match Windows, Light or Dark. Match Windows follows your system setting and changes with it, including automatic day/night switching. Light and Dark pin Lickety regardless of what Windows does.

START LICKETY WHEN I LOG IN — off by default. Turning it on registers Lickety with Windows startup; it appears in the tray a few moments after you sign in. If you're not sure whether it fired, the About screen can write a startup log.

OPEN THE MAIN WINDOW WHEN LICKETY STARTS — off by default. Lickety normally starts into the tray and stays out of the way, which is right once it's set up but awkward while you're still setting it up. Turn this on and this window opens too. It applies to EVERY start, not only the one at login — including when you run Lickety yourself.

RESTORE MY CLIPBOARD AFTER DICTATING — off by default, deliberately. Lickety's normal insertion method uses the clipboard and leaves your dictation on it; turning this on puts your old clipboard contents back afterwards. Restoring it can race the paste in some apps and put the OLD contents in, which is worse than losing a clipboard entry. Turn it on if you rely on your clipboard and you're not seeing that problem.

Every dictation is saved locally so you can find it again. Transcripts shows them newest first, with a search box over the list.

TYPE WHAT YOU'RE LOOKING FOR, THEN PRESS ENTER OR CLICK SEARCH. It does not search while you type, which is deliberate — with a large history, searching on every keystroke means most of the work is thrown away by the next one, and the box gets slow exactly when you have the most to search. Escape or CLEAR empties the box and puts the full list back.

Right-click a row and choose "Report an issue with this one". Use it when something came out wrong — the report captures what the recognizer produced AND what Lickety made of it, which is the pair needed to tell a mishearing from a formatting bug.

How many are kept is set under Settings → History (50,000 by default). Nothing leaves your machine; see Privacy.

Vocabulary is the single most valuable thing you can fill in, and it is what makes Lickety better than a generic dictation tool on YOUR words.

Add the words the recognizer has no reason to know: customer and vendor names, product codes, colleagues, internal jargon. They're fed to the recognizer BEFORE it listens, which biases it toward hearing them — it is not a find-and-replace after the fact.

THE NOTE COLUMN IS FOR YOU. It is never sent to the recognizer. Use it to record why a term is there so you can prune the list later.

WATCH THE CAPACITY METER. There is a hard limit on how much can be sent — roughly 224 tokens — and terms past it are SILENTLY IGNORED. The meter tells you where you stand. If you're near the top, remove the terms the recognizer already gets right; it knows ordinary English perfectly well and doesn't need 'invoice' or 'Microsoft'.

TWO TERMS THAT LOOK ALIKE CAN BLEND INTO A THIRD WORD THAT IS NEITHER. These terms are a hint to the recognizer, not a spelling list it matches against afterwards — so two that share a beginning or an ending compete instead of adding up. With both 'Inovisworks' and 'Innoapps' in the list, one dictation came out as 'Inoppsworks', which is neither word. If you see that, keep the term you actually say out loud and remove the other one.

WORTH KNOWING UP FRONT: this fixes the PREDICTABLE misses — the names it gets wrong every time. It will not make dictation perfect, and two things it cannot fix at all:

• Recognition occasionally invents something you didn't say, even from clean audio in a normal sentence. That's the technology, not your setup — every tool built on it does this.
• It only ever has the sound that reached the microphone. Mumbling, running words together or letting the end of a sentence trail off produces wrong text no matter what is in this list. Speak at a normal, even pace right through to the end.

TWO NEARLY-IDENTICAL TERMS CAN COMPETE INSTEAD OF ADDING UP. Because these are fed in as a hint rather than looked up afterwards, two terms that share a beginning and an ending can be blended into a third word that is neither of them — seeding both "Inovisworks" and "Innoapps" produced "Inoppsworks". If a term starts coming out worse after you added a similar one, keep the one you actually say and remove the other.

CHANGES TAKE EFFECT ON YOUR NEXT DICTATION. Edit the list and the recognizer is re-taught straight away — there is nothing to save, close or restart.

A term that still fails after being added is a different problem — see below.

Speech recognition sometimes hears a thinking-pause as the end of a sentence and capitalizes the next word by mistake — "the server was down So I restarted it", "remote desktop Sessions". This list fixes that.

HOW IT WORKS. If one of the listed words shows up capitalized in the MIDDLE of a sentence, it's put back to lowercase. That's all it does. It never adds a capital, never changes a word, and never touches a real capital at the start of a sentence — "So" starting a sentence stays "So".

A word that ISN'T on the list is left completely alone, so names, companies and products are never lowercased. That safety is the whole reason it's a list: only words you've approved are ever touched.

THE RULE FOR WHAT TO ADD — a word that, in YOUR writing, is more often lowercase than capitalized. "Because", "And", "With", "Correct" belong here. A word that is usually a name or a title does NOT: "Mark", "Bill", "Rose", "Sessions". Lowercasing a name is the worse error, because you read straight past it.

Enter the word in its CAPITALIZED form ("Because", not "because") — that's the form being corrected. The Note column is only for you.

Anything in your Vocabulary list is protected automatically; that list is your declaration of "these are proper nouns I care about".

The opposite of Auto-Lowercase. Sometimes a person's name arrives without its capital — "Met with pete, russell, and gary". This puts the capital back.

WHY IT HAPPENS. Speech recognition often capitalizes the opening few words of a dictation whether they need it or not, so Lickety puts a run of capitals at the start of a sentence back to normal. Three names in a row look exactly like that run — "Pete, Russell, and Gary" and "Sales, Service, and Support" are the same shape, and nothing in the words tells them apart. So Lickety learns names instead of guessing at them.

MOST NAMES ARE ALREADY BUILT IN. Thousands of common given names ship with Lickety and are not shown in the list — you do not have to add Pete or Jennifer. The list is for the names the built-in set does not know: unusual spellings, surnames you dictate often, and names from outside the built-in set.

COMPANY NAMES ARE BUILT IN TOO. Say "burger king" or "home depot" on its own and it comes back as Burger King and Home Depot. Hundreds of companies ship with Lickety, including ones that are two or three words long.

YOU CAN ADD A PHRASE, NOT JUST A WORD. "Pickleball Kingdom" or "Uncle's Family Kitchen" go in the list exactly as you would say them. This is where your own local businesses belong — the built-in list knows the national chains, not the restaurant down the road.

IT ONLY EVER ADDS A CAPITAL. It never lowercases anything, never changes a word, and never restyles a capital you already have — "McDonald" and "PDF" are left exactly as they are.

THE ONE THING TO KNOW BEFORE YOU ADD A NAME. Some names are also ordinary words: Bill, Mark, Grace, May, Art, Rose, Will. Lickety cannot tell which you meant, so adding "mark" also capitalizes "Mark it as paid". That is why those names are NOT built in. You can still add one — the panel warns you and then does what you asked. Add it if the person comes up more often than the word does.

THE SAME IS TRUE OF COMPANIES, AND IT IS WHY SOME BIG ONES ARE MISSING. Apple, Target, Gap, Subway, Square, Visa, Zoom and Slack are all ordinary English words, so they are NOT built in — otherwise "I ate an apple" and "mind the gap" would sprout capitals. Add one yourself if you dictate it far more often than the word, and the panel will warn you first.

TO TURN IT OFF, untick "Put capitals back on names" at the top of the panel. To stop one built-in name or company only, type it into "Built-in names to ignore".

CAPITALIZE MOVIE TITLES AFTER "THE MOVIE" — on by default. Say "we went to the movie jaws" and it lands as "the movie Jaws"; "my favorite film is the godfather" lands as "The Godfather". Thousands of well-known titles are built in.

IT ONLY LOOKS FOR A TITLE RIGHT AFTER THE WORD "MOVIE" OR "FILM" ("the movie…", "the film called…", "my favorite movie is…"), or in the sentence after one that ends in "movie", after "seen", "watched", "like" and similar: "We saw a great movie. Have you seen Jaws yet?" That gate is why "look it up", "I gave it to her" and "the heat was big today" are never touched, even though Up, Her, It, Heat and Big are all titles.

"THE MOVIE WAS GOOD" STAYS AS YOU SAID IT. After "is", "was" or "were", or after "seen / liked" in the next sentence, a ONE-word title that is also an everyday word — Good, Here, Up, It, Home, Big and the like — is read as a description, not a title. Say "the movie called Up", "the movie Up", or put it in quotes, and it is capitalized. Longer titles are not affected: "the movie was the godfather" still gives The Godfather.

ADDING TITLES OF YOUR OWN. Put them one per line in a file called MovieTitles.user.txt in the same folder as your settings file, with the capitalization you want. They are read the next time Lickety starts. There is no list on screen for this on purpose — it is a file, and one line per title.

Turn it off on the Auto-Uppercase screen.

Speech recognition sometimes hears a tiny pause before a trailing word and splits it into its OWN little sentence. You say "…it's a new file anyway" and it writes "…it's a new file. Anyway." This list glues that stray fragment back on.

HOW IT WORKS. If a listed entry appears ALONE as its own short sentence — up to three words — it is merged into the previous sentence in the form you specify. "…new file. Anyway." becomes "…new file, anyway."

It only fires when the entry stands completely alone between two sentence endings, AND only when you have listed it. Words that genuinely start a sentence — "Anyway, let's move on." — are left exactly as they are, and an unlisted phrase is never touched at all.

TWO COLUMNS: • WORD — the standalone word or phrase as it appears, capitalized ("Anyway", "For the disclaimer"). • REPLACE WITH — exactly what to put in its place, usually the lowercase word with a leading comma (", anyway"). The form is your choice, with or without the comma; the sentence's own ending period is kept either way.

WATCH THE LEADING SPACE — THIS IS THE ONE THAT CATCHES PEOPLE. Whatever you type in REPLACE WITH is put in EXACTLY as you typed it, with nothing added. So if you leave off the leading comma AND the leading space, the replacement runs straight into the previous word: "…close the log. File entry." with a replacement of "file entry" gives you "…close the logfile entry." Type " file entry" — with the space — and you get "…close the log file entry."

Nothing is inserted for you on purpose, because joining the two words with no space is sometimes exactly what you want. A leading comma already carries its own space, which is why the usual ", anyway" form never shows this.

You are part-way through a sentence, you stop — interrupted, or still deciding how to finish — and then come back with a two or three word dictation to complete the thought. The recogniser writes every dictation as its own sentence and title-cases the short ones, so what lands in the middle of your paragraph is "Price List Control" rather than "price list control".

LOWERCASE A DICTATION OF THREE WORDS OR FEWER — on by default. Three words or fewer, the whole dictation goes lower-case.

THE RULE IS FLAT ON PURPOSE. It does not look at what you typed before, and it does not try to work out whether you are starting a sentence or continuing one. Over Remote Desktop it cannot see the cursor at all, and that is exactly where this happens most — so a rule that depended on seeing it would help you least where you need it most.

THREE THINGS IT NEVER LOWERCASES: words in capitals (US, PDF, VPN), words with a capital inside them (iPhone, eBay), and anything in your Vocabulary list.

THE TRADE. A short dictation that really is a proper noun — "Home Depot", a three-word book title — comes out lower-case. If you dictate those on their own often, add them to Vocabulary or turn this off.

Longer dictations are untouched: four words and up have their own rule.

The recognizer guesses at how to write numbers, and its guesses suit prose rather than data. This screen takes the guessing back.

ENABLE NUMBER FORMATTING — the master switch. It also strips stray dashes the recognizer inserts into long digit runs ('12345-6').

NUMBER TYPES — each type has a TRIGGER PHRASE and a FORMAT TEMPLATE. Say the trigger, then the digits, and the digits are written to the template. A phone template of 469-326-9458 turns ten spoken digits into that shape. Leave a template blank to get plain digits with nothing added.

EXPAND ABBREVIATED NUMBER SEQUENCES — on by default. 'seven point one, two and three' becomes '7.1, 7.2, 7.3'. Turn it off if you dictate genuine short lists of numbers.

USE THE OXFORD COMMA IN LISTS — off by default. Whether a list is written 'A, B, and C' or 'A, B and C'.

A NOTE ON SPOKEN DIGITS. Reading a long number aloud digit by digit is the hardest thing you can ask of a recognizer — it may add or drop one. Check long numbers.

WRITE SPOKEN MONEY AS DIGITS — on by default. Say an amount and it is written as digits: "three hundred and thirty eight dollars" becomes $338, and "five dollars and ninety nine cents" becomes $5.99.

MOST OF THE TIME THERE IS NOTHING TO FIX. The recogniser usually writes $338 on its own. Every so often — typically when you are speaking quickly with two amounts close together — it writes the words out instead, and there is no way to predict which you will get. This makes the answer the same either way.

IT ONLY FIRES ON A NUMBER FOLLOWED BY "dollars" OR "cents". "The dollar store", "the dollar sign" and "it cost a few dollars" are all left alone, because none of them has an amount in front of the word.

"A MILLION DOLLAR IDEA" IS NOT AN AMOUNT, and it is left alone. English marks the difference itself: an amount uses the plural ("it costs five dollars"), while a description uses the singular ("a five dollar bill", "a million dollar idea"). Nobody writes "a five dollars bill", so this needs no guesswork.

LARGE AMOUNTS: DIGITS OR WORDS, YOUR CHOICE. Formatting → Number Formatting → WRITE AMOUNTS OF AT LEAST sets where the switch happens. The default is a million, so "two million dollars" gives $2 million while "three thousand dollars" still gives $3,000. Set it to a billion and millions stay in digits; set it to never and everything does.

The point of the setting is that huge amounts stop being readable as digits somewhere, and where that is depends on what you write — $100,000,000,000,000 is a number you count the digits of rather than read.

ONLY WHOLE AMOUNTS, HALVES AND QUARTERS become words. $2,000,000 gives "$2 million" and $1,500,000,000 gives "$1.5 billion", but $2,175,000 stays in digits at every setting, because "$2.175 million" is an improvement on nothing. Those are the fractions English has words for, which is exactly what makes them readable.

WHATEVER YOU PICK, YOU GET IT EVERY TIME. The recogniser writes the same amount as "$2 million" some days and "$2,000,000" others, with no way to predict which. All three spellings are read back to a number and written out your way.

"$12.00" OR "$12"? The recogniser writes it both ways depending on the day. SHOW ".00" ON WHOLE AMOUNTS in Number Formatting settles it: off — the default — always gives $12, and on always gives $12.00, whichever way the recogniser wrote it. Cents you actually said are never touched, so $12.50 stays $12.50 either way, and an amount written in words ("$2 million") never gets .00.

WRITING THE WORD "DOLLARS". Normally the word is absorbed into the "$", so "one point five billion dollars" gives $1.5 billion. If you would rather see it — $1.5 billion dollars — turn on WRITE "DOLLARS" AFTER AN AMOUNT in Number Formatting.

EVERY dollar amount carries it, not just the ones where your spoken word happened to survive. The recognizer writes "five dollars" some of the time and "$10,000" other times for the same spoken shape, so a switch that only kept the word when it was still there would give it to you on some lines and not others — which is the unpredictability this whole screen exists to remove. Where you used your own word it is kept: say "bucks" and you get "bucks".

It stays off when cents are folded in, because "$12.50 dollars" is wrong in a way "$12 dollars" is not.

ONE THING IT WILL NOT DO. If you say two amounts and only put "dollars" after the first — "the old price is three hundred and thirty eight dollars, the new one is three hundred and ninety two" — only the first converts. Carrying the word across would mean guessing that a bare number is money, and a guess about your prices is a bad place to start. Say "dollars" after each one.

Turn it off in Number Formatting if you would rather keep what the recogniser gave you.

WRITE SPOKEN ARITHMETIC AS SYMBOLS — on by default. Say a sum and the operator words become their symbols: "187 times 12" gives 187 × 12, "25 plus 30 divided by 2" gives 25 + 30 ÷ 2, and "10 take away 4" gives 10 - 4. The words it knows: plus, add, added to; minus, subtract, less, take away; times, multiplied by, multiply by; divided by — and a lone letter x between two numbers, which is how the recogniser sometimes writes "times".

"OVER" IS LEFT AS A WORD. "Sales grew 15% over 2024" and "400 over 3 days" have a number on both sides and are not sums, so "over" never becomes ÷. Say "divided by" for a sum.

IT ONLY FIRES WITH A NUMBER ON BOTH SIDES OF THE WORD, and that is the whole safeguard. "Three times a day", "take 5 minus the tax" and "I was over 10 minutes late" all have a word rather than a number on one side, so they are left exactly as you said them. Ordinary writing is never touched.

A SUM ON ITS OWN LOSES ITS FULL STOP. Tidy-up ends every dictation with a period, and "187 × 12." is not what a calculator wants. When the whole dictation is nothing but numbers and operators, the period is dropped. A sum inside a sentence keeps the sentence's punctuation.

START WITH "EQUALS" FOR A SPREADSHEET FORMULA. Say "equals 187 times 12" and you get =187*12 — the leading equal sign Excel needs, no spaces, no thousands commas, and the keyboard operators * and / that a formula (and Windows Calculator) accept, since neither takes × or ÷. Only a dictation that STARTS with "equals" and is nothing but a sum does this; it works the same in every app, because "equals five plus three" is only ever a formula.

IN A SPREADSHEET, START WITH THE OPERATOR TO CONTINUE A FORMULA. Double-click a cell that already holds =1250*3.5, say "times 2" (or "times two" — a number word counts when the whole dictation is a sum), and *2 is typed at the end — no period. That only happens in a spreadsheet, only when the operator is the first word, and only when the whole dictation is operators and numbers; "times" inside a sentence typed into a cell is still a word.

A PAUSE INSIDE A SUM IS FINE. The recogniser sometimes writes your pause as a comma ("10, take away 4") or runs two words together ("takeaway"); in a dictation that is nothing but a sum both are understood. Inside a sentence a comma is yours and is kept.

WHAT IS DELIBERATELY LEFT AS WORDS. "Equals" at the END of a sum — a calculator treats the word as its own button, and = in the middle of a sentence looks wrong. "Of" — "20 percent of 80" reads fine as it is. "To the power of", "squared" and "square root of" — no symbol makes them clearer. "Plus or minus" is a phrase, not a sum, and stays one.

MINUS IS THE KEYBOARD HYPHEN, not the typographer's minus sign, because the hyphen is what every calculator, spreadsheet and terminal accepts as subtraction. Multiply and divide are the real × and ÷ symbols, never a letter x.

Turn it off in Formatting → Number Formatting if you would rather keep the words.

Say it the way you would read it out to someone: "paul at great scott labs dot co". Lickety writes paul@greatscottlabs.co — lower-case, no spaces, the way addresses are written. Always on — there is no switch to find: an address needs a known ending (dot com, dot co…) before the rule does anything, so a sentence that is not an address is never touched. Web addresses work the same way — see WEB ADDRESSES below.

WHAT IT NEEDS. Three parts, in order: the name; the word "at" (or an @ the recogniser already wrote); the company — up to four words — and then "dot" with a common ending: com, net, org, co, io, us, gov, edu and a dozen more. The ending is what makes it an address: "meet Paul at Acme" is a sentence and is left alone, and so is "at the office dot later", because "later" is not an ending.

NAMES WITH PARTS. "paul dot scott at acme dot com" → paul.scott@acme.com. "first underscore last" and "first dash last" write the symbol. A company with parts works the same way: "sales at mail dot acme dot com" → sales@mail.acme.com.

CAPITALS AND SPACES. The recogniser capitalises company names and leaves spaces between the words — "Paul at Great Scott Labs.co" — and a Replacement of yours may capitalise them too. An address is case-insensitive, so Lickety lower-cases the whole thing and closes the spaces. If you want capitals kept in an address, say it, and I will make it a switch.

ONE SENTENCE, SEVERAL ADDRESSES. Each is found on its own: "I have paul at acme dot co, support at acme dot co and junk at acme dot co set up" writes all three. Say "comma" between them if you want the commas; the recogniser usually leaves them out.

WEB ADDRESSES WORK THE SAME WAY. "www dot great scott labs dot co" → www.greatscottlabs.co; "greatscottlabs dot co slash pricing" → greatscottlabs.co/pricing; "https colon slash slash greatscottlabs dot co" → https://greatscottlabs.co; "192 dot 168 dot 1 dot 1" → 192.168.1.1. "www" or a scheme in front lets the company be several words; without one, a company of several words is read as the address only when the recogniser capitalised it as a name ("Great Scott Labs.co"), so "visit greatscottlabs dot co" does not swallow the word "visit". An address the recogniser already wrote whole is left exactly as it is.

YOUR OWN ADDRESSES, FASTER. For the two or three you use every day, a QuickText — "my email" → paul@greatscottlabs.co — is one phrase and never mishears a letter.

"Ship it to 123 Main Street, Suite 400" often comes back with SWEET 400, because the two words sound almost identical and the recogniser has no idea it is reading an address. Lickety puts it right: Formatting → Addresses → FIX "SWEET 400" TO "SUITE 400", on by default.

IT LOOKS AT WHAT FOLLOWS, not at the sentence around it. A number — on its own or with a letter — is what makes it an address: Sweet 400, Sweet 4B, Sweet 17-A. That shape does not follow the ordinary adjective, so "that is a sweet deal" and "sweet and sour chicken" are never touched, and you do not have to say the word "address" or any other magic phrase for it to work.

A CAPITAL LETTER THAT ENDS THE PHRASE counts too — "Suite A," and "Suite B." A lower-case one does not, because "it was sweet a little too much" has to survive.

NUMBERS THAT STAY "SWEET" is the short list underneath, seeded with 16 for the birthday and the basketball. Edit it BOTH ways — if your own office is Suite 16, delete that row and it will be corrected like any other.

WHY NOT JUST ADD "SUITE" TO VOCABULARY? Because that is the usual answer and it is the wrong one here. Vocabulary nudges the recogniser toward a word across everything you say, so teaching it "suite" would put the real word "sweet" at risk in every other sentence. Use Vocabulary for a word you always mean; use this for a word that depends on where it sits.

DETECT SPOKEN DATES — on by default — turns something said as a date into a written one — 'six sixteen twenty twenty six' becomes 6/16/2026.

WRITE DATES AS chooses the shape: 6/16/2026, 6/16/26, or June 16, 2026.

DATE TRIGGER PHRASES are the words that mark what follows as a date — 'the date is', 'dated', 'on the date of'. A trigger makes detection far more reliable, because a bare run of numbers is ambiguous and always will be.

PHRASES FOR SEVERAL DATES IN ONE SENTENCE are a second, separate list. A phrase from the first list reads the date you say NEXT. A phrase from this one claims every date in the rest of the sentence, however many words come in between — "the dates that I have open for an appointment are 731 26 or 8126" converts both.

They are two lists rather than one setting because reaching past several words is only safe once you have said there are several dates coming. "The date is" promises the opposite, and stretching it would let any number later in the sentence be read as a date. A bare "the dates" is deliberately not included for the same reason — "I checked the dates and 12345 items were missing" would turn a part number into a date. Add it if you want it, knowing that.

KEEP THE TRIGGER PHRASE IN THE TEXT — off by default here. Off, a phrase from the list is treated as a marker you said to introduce a date, and it is replaced: "the date is 73126" gives you "7/31/2026". Turn it ON when the phrase is part of a sentence you actually want to write, and you get "The date is 7/31/2026" instead.

This only affects a listed phrase matching on its own. "Tomorrow's date is 8126" keeps its words either way, and so does anything caught by the several-dates list.

Turn detection off entirely if you dictate numbers that keep being read as dates.

TWO WAYS, and they cover different moments.

CAPS LOCK — on by default. Switch Caps Lock on before you press the hotkey and the whole dictation arrives in capitals. Switch it off and the next one is normal. The light on your keyboard is the reminder, which is the reason this is the better of the two for a whole line.

Caps Lock on, then "quarterly results" -> QUARTERLY RESULTS

IT IS READ WHEN YOU PRESS THE HOTKEY, not when the text is typed. Lickety types exact characters rather than pressing keys, so Windows never applies Caps Lock on its behalf — the app reads the key itself, at the moment you decided.

THE SPOKEN PHRASES — on by default, and FIXED. Say "mode uppercase start", and everything after it is capitalised. Say "mode uppercase stop" to go back to normal:

"the heading is mode uppercase start quarterly results mode uppercase stop and then the body" -> the heading is QUARTERLY RESULTS and then the body

THE STOP IS OPTIONAL. Leave it out and the rest of the dictation is capitalised, which is usually what you want when the capitals come last.

THE PHRASES NEVER APPEAR IN YOUR TEXT, and they cannot be edited. They were chosen because nobody says them by accident — a shorter pair like "mode up" would fire on ordinary sentences such as "take maintenance mode down at six".

BOTH SWITCHES ARE IN Formatting -> Uppercase, and they are independent — turning one off leaves the other working. They are labelled "Dictate in capitals when Caps Lock is on" and "Obey \"mode uppercase start\" and \"mode uppercase stop\"".

DETECT SPOKEN TIMES — on by default. Whisper writes a spoken "ten thirty" as 10.30, 10 30 or 10-30 — almost never 10:30 — and you would never say "colon" in the middle of a time, so nothing else can put it right.

Say one of the TRIGGER PHRASES and the time straight after it is rewritten:

"the meeting is at ten thirty" -> the meeting is at 10:30 "the restaurant is open from ten thirty a m to nine p m" -> the restaurant is open from 10:30am to 9pm

A RANGE CARRIES ACROSS. Once one time has been converted, a following "to", "until" or "through" plus another time is converted as well — that is how the opening-hours example works, with no rule of its own.

WHAT IT WILL NOT TOUCH, deliberately: • a bare hour — "the meeting is at 10" stays as it is, because turning that into 10:00 would be guessing • "noon", "midnight", "half past ten" — already correct English • anything that is not directly after a trigger phrase

ABOUT A HUNDRED PHRASES SHIP WITH LICKETY — "meeting is at", "finishing by", "due by", "leaves at", "remind me at" and so on. Number Formatting -> Times -> BUILT-IN PHRASES opens the list; untick any phrase to stop it, and it stays off until you tick it again. New ones arrive with each update by themselves. YOUR OWN PHRASES go in the grid under that button, on top of the built-in ones. "___" in a phrase stands for one to four words, so "meet for ___ at" covers "meet for lunch at" and "meet for the quarterly review at".

A PHRASE THAT BEGINS WITH "#" MEANS THE TIME COMES BEFORE THE WORD rather than after it. That is the only thing the "#" does, and it is why one list covers both shapes instead of you keeping two:

a plain phrase, time after: "the meeting is at two thirty" -> the meeting is at 2:30 a "#" phrase, time before: "the 230 meeting request came in" -> the 2:30 meeting request came in

Among the ones that ship this way are "# meeting", "# appointment", "# call", "# flight", "# class" and "# shift", and you write your own the same way. THE "#" IS NEVER MATCHED AGAINST YOUR SPEECH and never appears in your text — it marks where the time sits and nothing else.

THE SAME REFUSALS APPLY to a "#" phrase. "The 3 meeting" is a bare hour and is left alone. "The 2026 meeting" is not a clock reading and stays a year. "12 meetings" does not match "# meeting", because the whole word has to match. And it will not reach back across a full stop or a comma, so "I called at 4. Meeting notes follow" leaves the 4 alone.

12-HOUR OR 24-HOUR follows your Windows regional setting rather than a switch here — on a 24-hour machine you get 22:30 and the am/pm style below does nothing. How am/pm is written IS a setting: 10:30am, 10:30 AM or 10:30 a.m.

You can bind a mouse button to COPY or RE-INSERT your last dictation. Settings → Mouse button.

IT IS BUILT FOR REMOTE DESKTOP. Mouse events are handled on your own machine before the RDP client forwards anything, so a mouse button gets through in a full-screen session where a keyboard shortcut may not.

YOU BIND IT BY PRESSING IT. Click "Click a mouse button…" and then press the button you want. There is no list to choose from on purpose — "Back" and "Mouse 4" are not the same thing on every mouse, and pressing the button also proves Windows can see it.

IF PRESSING DOES NOTHING, your mouse's own software has claimed that button before Windows sees it. Logitech Options, Razer Synapse and similar do this. No program can reach a button in that state — change it in your mouse software.

LEFT AND RIGHT CANNOT BE BOUND. If you bound left-click you could not click your way back to this setting to undo it. The two thumb buttons and the wheel click are offered; losing one of those is an annoyance, not a lockout.

WHILE A BUTTON IS BOUND IT STOPS DOING ITS USUAL JOB — no more Back, Forward or open-in-new-tab on that button. Clear the binding and it goes straight back.

IT INSERTS WHATEVER TRANSCRIPTS IS POINTING AT. Normally that is your last dictation. But if you press COPY on any row in Transcripts, that row becomes the one the button will insert — the button says "Ready to insert" for a moment so you can see it took. Your next dictation takes over again.

THAT IS THE ANSWER TO A REMOTE DESKTOP DEAD END. Inside an RDP session Ctrl+V pastes the REMOTE machine's clipboard, not yours, so copying an older transcript here and pasting it across simply does not work. The mouse button goes around the clipboard entirely.

TWO THINGS IT CAN DO. RE-INSERT IT AT THE CURSOR is the default and needs nothing else from you — press the button, the text appears. COPY IT puts the text on the clipboard instead and leaves you to press Ctrl+V.

WHY WOULD YOU EVER PICK COPY, THEN? For the one case where re-inserting is the risk: if an insert has just failed or come out garbled in an app, sending the text back down the same path may do the same thing again. The clipboard cannot be garbled, so copy is the recovery option — worth having, and worth choosing deliberately rather than being handed by default.

Lickety reopens where you last left it, including on a second monitor. There is one case where that goes wrong, and it depends on HOW the other monitor is off.

UNPLUGGED (video cable out) — Lickety comes back on the monitor you still have. Windows drops the display, so the old position is genuinely nowhere and the window is re-centred.

POWERED OFF (still plugged in) — it does NOT come back, and it can't. Windows keeps a switched-off-but-connected monitor in the desktop, so as far as every check goes the window is on a real screen. It is — a real screen that happens to be dark. There is no way for a program to ask whether a monitor is actually lit.

TWO WAYS TO GET IT BACK:

• Right-click the Lickety tray icon and choose OPEN ON THE MAIN MONITOR. This asks nothing and checks nothing — it just puts the window in the middle of your primary screen.

• Or use Windows itself: with the window focused, press Win+Shift+Left or Win+Shift+Right to move it to the next monitor. This works for any program, not just this one.

WHEN LICKETY CANNOT TELL WHAT TO REPLACE. A correction that starts with a plain verb and nothing else to anchor it — "send it to Bob, scratch that, cancel it" — could mean two different things, so Lickety leaves your words exactly as you said them and adds ** Scratch that can't determine the exact change ** at the end. The asterisks are there so it is obvious that nothing was changed. Fix the sentence by hand, or say it again with something to anchor the correction — a name, a number, a day. Corrections with a number ("order 500, scratch that, order 300"), a capitalised name ("in Dallas, scratch that, Fort Worth") or a day always resolve.

RECOGNISE SCRATCH THAT AS A COMMAND — off by default. Say "scratch that" (or "strike that") and Lickety treats it as a command rather than words to type.

IT'S OFF UNTIL YOU TURN IT ON. Formatting → "Scratch that".

THREE SHAPES, AND THEY DO DIFFERENT THINGS.

• WITH NOTHING AFTER IT, it removes the whole SENTENCE in front of it. Say "Please order me a sandwich. Scratch that" and nothing is typed — the entire sentence goes, not just the last few words. An EARLIER sentence in the same dictation is kept: "I'll be in at nine. Please order me a sandwich. Scratch that" leaves you "I'll be in at nine."

• IN THE MIDDLE OF A SENTENCE, it corrects what you just said: "I need to order 500 widgets, scratch that, 300 widgets" gives you "I need to order 300 widgets."

• "SCRATCH ALL THAT." (or "scratch that all.") THROWS AWAY EVERYTHING BEFORE IT and keeps what follows. It is for a false start: "I'm testing something with dictation, scratch all that. I'm going to go ride my bike." gives you "I'm going to go ride my bike." It fires only when the phrase ends a sentence — the next thing after it is a full stop. "Scratch all that, and then…" or "scratch all that nonsense" are words and are typed. Often the recogniser drops a false start by itself before Lickety ever sees it; this is for the times it doesn't.

IT ONLY CORRECTS WHEN IT IS SURE. When two readings of your sentence disagree it changes nothing at all and leaves a note in the text saying so. Nothing you said is ever removed on a guess — a visible refusal to correct is a smaller problem than a wrong correction you don't notice.

ORDINARY ENGLISH IS LEFT ALONE. "I need to scratch that itch" and "we should strike that clause from the contract" are not commands and are not touched. What makes the difference is a pause or a comma before the phrase — without one it is just words in a sentence.

DAYS AND MONTHS CORRECT IN PLACE. "Let's meet next Tuesday for lunch, scratch that, Wednesday" changes the Tuesday, wherever it sits in the sentence.

ENABLE LIST PHRASES — on by default. Say one of these phrases and the rest becomes a list, split at the natural pauses you speak. It can start the dictation or begin a later sentence — anything said before it is kept as ordinary text above the list.

"Here's my to-do list: walk the dog, feed the dog, bathe the dog"

Each row is a phrase, the STYLE it produces (bulleted or numbered), and whether to KEEP THE PHRASE as a heading above the list or drop it.

Nothing becomes a list unless you asked. Say anything else and your text is left exactly as you said it — Lickety never decides on its own that you meant a list.

A spoken line break ENDS the list — but if what follows starts with another list phrase, it becomes a SECOND list rather than ordinary text. Two lists in one dictation get a blank line between them, whether or not you asked for one.

IF YOUR PHRASE ENDS IN "of" AND YOU SAY A COUNT, THE COUNT JOINS THE HEADING. "Here's a list of three items: ..." keeps "of three items" in the heading instead of making it the first bullet. A count is a number plus a word like items or things; anything else after "of" is content and stays as your first item, so "here's a list of groceries" still lists the groceries.

Say one of these phrases and you get a real line break instead of the words.

The shipped phrases are 'line break', 'paragraph break' and 'new paragraph'. BLANK LINE decides whether the phrase leaves an empty line after it (a paragraph) or just moves to the next line.

WHY NOT 'new line'? Because it is ordinary English — 'we added a new line to the production floor' — and a break phrase that fires inside a sentence silently destroys text you'll never see was removed. You can add 'new line' or 'next line' yourself; just know what you're accepting.

THE FIRST WORD AFTER A BREAK IS CAPITALISED, always. There is no setting for this. The recogniser writes the word after your pause in lower case, because mid-sentence that is what it heard — and it is not even consistent about it, so the same phrase can come back capitalised one minute and not the next. Once the break turns the words in front of it into the end of a line, a lower-case word there reads as a mistake, and nothing else goes back to fix it. Asking for a new line is asking for a new line.

A break phrase only counts as a command at a natural boundary: the start of what you said, after a pause, or straight after a symbol or a part number. Anywhere else mid-sentence, it stays as words.

Say a trigger phrase — 'compose email' by default — and what follows is shaped like an email rather than a paragraph: a greeting on its own line, the body, and a sign-off.

"Compose email. Eric, let's meet Thursday to discuss the issue. Thanks."

YOUR NAME is added after a sign-off word when you didn't say one, so 'Thanks' becomes 'Thanks,' and your name on the next line. Leave it blank and nothing is added.

SIGN-OFF WORDS is the list that counts as an ending — thanks, regards, best, and so on. It's a LOOKUP, not a guess about sentence shape, which is why it never mistakes an ordinary closing sentence for a sign-off.

TREAT A NAME AT THE START AS A GREETING — turn this off if you routinely begin emails with something other than a name.

OTHER WAYS YOU SIGN OFF holds YOUR own names, and it is the only list that can end an email. Don't put people you write TO in it — that would sign your message with their name.

A NAME AT THE START DOESN'T NEED A LIST. Any capitalised word followed by a comma opens the email, whether Lickety has heard of the person or not. If the recogniser writes the name in lower case, or gets it wrong — 'Mike' heard as 'mic' — add a Replacement for it. That fixes the spelling and the capital together, and it runs long before the email is assembled.

'Ten to fifteen percent' should be written '10% to 15%', with the unit on both numbers, not '10 to 15%'. This table is which units get that treatment and how.

Each row is the SYMBOL, the SPOKEN WORD that means it, and whether the symbol goes AFTER the number (10% but $10).

RANGE CONNECTORS are the words that join the two numbers — 'to', 'through', 'or', 'and'. Remove one if it's firing on something that isn't a range.

Say "the file name is" (or "create filename", or any phrase you add) and everything after it is written as a NAME rather than a sentence — no commas, no full stop, no question mark.

"the file name is priceless documentation updates august first twenty twenty six" → The file name is Priceless Documentation Updates 08-01-2026.

KEEP THE TRIGGER PHRASE IN THE TEXT — on by default here. The phrase stays in the text, because most of the time you are writing a sentence ABOUT a file. Turn it OFF when you are dictating straight into a Save box and want the name on its own:

"create filename priceless documentation updates august first twenty twenty six" → Priceless Documentation Updates 08-01-2026

WHY THIS IS A SEPARATE MODE. A file name and a sentence want opposite answers to the same questions. A sentence wants commas and a full stop; a name wants neither. A sentence wants ordinary capitalisation; a name often wants Title Case. And a date in a sentence is written with slashes, which Windows will not accept in a file name at all. None of that can be settled by one rule, so you declare it instead.

DETECT DICTATED FILE NAMES — on by default — is the master switch for everything in this topic. Turn it off and a spoken file name is written as ordinary words.

TITLE CASE FILE NAMES — on by default. Short joining words (of, the, for, with) stay lower-case unless they start or end the name, which is what title case means everywhere else. An ALL-CAPS word is left alone, so UPC and PDF survive.

DATES IN FILE NAMES have their own format setting, separate from the one under Number Formatting: 08-01-2026, 2026-08-01 (which sorts correctly in a folder), or August 1, 2026.

SPOKEN SYMBOLS STILL WORK. Say "dash" for a dash and "underscore" for an underscore, exactly as you would anywhere else.

The phrases are all several words long on purpose, because this trigger claims everything you say after it. You can add your own — keep them specific.

Say a path and get a real one:

"T colon backslash inventory backslash data backslash inventory data dot mdb" becomes T:\Inventory\data\inventorydata.mdb

The spaces the recognizer puts around the separators are removed, and the first folder gets a capital.

THE EXTENSION TABLE decides whether the spaces INSIDE a filename survive. A database or a script almost never has spaces in its name, so those close up. A Word or Excel document usually does, so those keep them. An extension that isn't listed keeps its spaces — leaving a space in is a smaller, more visible mistake than welding two words together. Folder names always keep their spaces, so C:\Program Files\ is safe either way.

CAPITALS IN A FILE NAME. "Capitalize the first letter of each word in a file name" turns "help_refactor.md" into "Help_Refactor.md". A SPACE ends the file name, so "help refactor.md" only capitalizes "Refactor" — the earlier word belongs to your sentence, not to the name. A word that already contains a capital is left exactly as it is, so README.md and MyNotes.docx are safe. The extension itself is always written in lower case.

MISHEARD DRIVE LETTERS ARE HANDLED. The recognizer writes a spoken 'C' as 'See' and a 'T' as 'Tea'; those are understood in the drive position. It also sometimes writes a forward slash for a spoken 'backslash' — either one works, and you always get a backslash out.

AND SO IS A MISHEARD "COLON". This is the one that breaks a path completely, because the colon is what tells Lickety a path has started at all. The recognizer writes the spoken word as a name — 'W. Cohen/Reports/' or 'W, Colin, \Reports\' — and the whole path arrives as ordinary words.

The field on the Formatting screen, "Words Whisper writes instead of 'colon' in a drive path", is that list. Add whatever yours turns into. It is only ever consulted between a SINGLE drive letter and a run of slashes, so a colleague called Colin is safe everywhere else in your writing — the position does the work, not the word.

TWO THINGS IT DELIBERATELY WILL NOT DO. If the recognizer drops the colon entirely ('W Reports/Archive/'), nothing is repaired — guessing there would silently eat 'Reports' as though it were the missing colon, and you would get a path that looked perfectly correct. And 'A' and 'I' are never treated as drive letters in this case, because they are ordinary English words and a sentence should not become a path. Saying 'A colon backslash' still works normally.

YOU DO NOT NEED TO SAY "PERIOD" OR "COMMA". The recognizer already writes those four marks — period, comma, question mark and exclamation — from the way you speak. Saying the words out loud is not just unnecessary, it is the one thing most likely to go wrong: the recognizer has to get the WORD right too, and "comma" comes back as "Kama" and "period" as "Pyrion" often enough to notice. Just say the sentence and pause where the punctuation goes.

IF YOU SAY THEM ANYWAY, Lickety drops the word and keeps the mark — including when the mark ends up inside a quotation. It only does that where the recognizer wrote the matching mark, so "we had a good period, all things considered" and "I said a comma" are safe.

Everything OTHER than those four arrives as the word, and that is what this table covers. It ships filled in.

Say 'backslash' and get \. Say 'open paren' and get (. Say 'underscore' and get _.

IT RUNS AFTER YOUR REPLACEMENTS, so anything you built yourself still wins. A shipped default can never override your own tuning.

QUOTATION MARKS. Say "open quote" and "close quote" (or "unquote", or "end quote") and you get a straight " at each end, closed up against the words rather than with a gap.

You often will not need them. The recognizer writes quotation marks by itself when it hears quoted speech — and when it does BOTH, writing the mark and leaving your spoken word in the text, Lickety drops the word and keeps the mark rather than giving you two. That applies to every row in this table, not just quotes.

Bare "quote" is NOT included, for the same reason as "dash": it is an ordinary word ("send me a quote for the work"). Add it yourself if you want it, knowing that trade.

DELIBERATELY NOT INCLUDED: bare "slash" and bare "dash", which are ordinary words ("slash the budget", "a dash of salt") — "forward slash" is here instead. And mishearings like "Colin" for "colon", which would rewrite a colleague's name. Add either yourself under Replacements if you want them.

ONLY AS PUNCTUATION is for words that are also ordinary words. 'Colon' is ticked, because it is also an organ: with the tick it only becomes ':' where it cannot be anything else — when you say 'space' after it, before a backslash, with a number either side, at the end of what you said, or straight after a list phrase. 'Something about the colon' is left alone. Tick it on any row of your own that's causing the same trouble.

IN THE MIDDLE OF A SENTENCE, SAY 'COLON SPACE'. 'The subject is colon space quarterly review' gives 'The subject is: quarterly review'. That is the dependable way.

A PAUSE EITHER SIDE OFTEN WORKS TOO. Say 'Lickety', pause, 'colon', pause, 'Voice Typing' and you get 'Lickety: Voice Typing'. It works when the recognizer writes your pauses down, as commas or as dots, and it does not always write them. If 'colon' comes out as a word, that is what happened: say 'colon space' instead. A pause before it is enough when the next word starts with a capital, as a name does.

IT STAYS A WORD WHEN YOU ARE TALKING ABOUT IT: after 'the' or 'a', after 'actual', 'the word', 'say' or 'called' ('change it to an actual colon'), in a list of marks ('a dash, colon, hyphen'), and when 'and' or 'or' comes next ('the stomach, colon, and liver'). A longer list with 'colon' in the middle, such as 'the stomach, colon, liver and kidneys', is read as a pause either side and gets the mark. If you dictate lists like that, delete the row.

A WORD AFTER 'the' or 'a' IS NEVER CONVERTED. 'The dollar sign works' stays as words; 'dollar sign 338' gives $338.

SOME WORDS THE RECOGNIZER SIMPLY CAN'T HEAR. 'Paren' comes back as 'parenne', 'parentacy', even 'friend'. The known mishearings are in the table, but if a symbol keeps failing, try a word that sounds less like everything else — 'bracket' is far more distinctive than 'paren', and you can point those rows at whichever character you want.

INPUT DEVICE — which microphone to record from. 'System default' follows whatever Windows is set to. Pick a specific one if you have several and Windows keeps choosing the wrong one.

The microphone is the single most common cause of bad transcription. A headset mic close to your mouth beats a good laptop mic across a desk, every time.

CHECK THE LEVEL — press "Test my microphone" and say a sentence normally. The bar shows what the microphone is ACTUALLY hearing, which is the only way to settle "is this thing picking me up" without dictating and reading the result.

• Nothing moving — wrong device, muted, or Windows is not letting apps use it (WINDOWS Settings → Privacy → Microphone — Microsoft's Settings app, not this one).
• Barely moving — move closer, or raise the input level in Windows sound settings. Quiet audio transcribes badly.
• Pinned at the top — too loud. Speech that clips loses the detail the recognizer needs.

It only listens while you have pressed Test, and it hands the microphone straight back the moment a dictation starts — so it can never be the reason a dictation went wrong. The level it warns about as silent is the same one dictation uses to refuse a capture outright.

WHICH MODEL FITS THIS MICROPHONE — press "Run the ten-second test" and read the three sentences on screen at your normal volume. Lickety ships two speech models: the faster one is accurate on a good microphone, and the more careful one recovers most of what a quiet or noisy microphone loses. The test measures how loud your voice arrives, how far above the microphone's own hiss and the room it sits, and whether the microphone is clipping, then says what it found and picks the model. The line under it always says which model is in use and who chose it — the test, you on the Settings screen, or the default. The first time Lickety starts it opens on this screen until you run the test or press "Not now". Nothing is recorded and nothing is sent anywhere.

PLAY A CHIME — a short sound when recording starts and stops, for when you're not looking at the screen. Off by default; the volume slider is separate and starts at 25%.

AUDIO WHILE RECORDING — leave other audio alone, lower it, or mute it while you dictate. Useful if you dictate with music or a call running.

TIDY UP TEXT is a set of small deterministic cleanups applied to what the recognizer produced. No AI, no rewriting — it never changes your words, only how they're written.

• capitalizes the first word
• fixes capitals the recognizer added at a pause mid-sentence
• collapses stutters and repetition artifacts
• removes sound annotations like [BLANK_AUDIO] and (keyboard clicking) — but keeps anything else in parentheses, such as an area code or an aside
• tidies spacing around punctuation

ADD A PERIOD AT THE END OF EACH DICTATION and the two TITLE-CASE switches are separate, under it, so you can keep the cleanup and drop any of them.

THE TWO TITLE-CASE SWITCHES — FIX TITLE-CASE WHEN THE WHOLE PHRASE IS CAPITALIZED and FIX TITLE-CASE WHEN ONLY THE OPENING IS CAPITALIZED, both on by default. They are separate because they fix opposite things and one of them can cost you a name.

The recognizer sometimes writes a dictation as if it were a heading. When it does that to the WHOLE thing — "The End of Another Word" — the first switch puts it back to sentence case. That one is safe: every word being capitalized is not something ordinary writing does.

When it does it to only the OPENING and then writes normally — "The Dollar Sign Works, that's the big one." — the second switch fixes that. **This is the one worth knowing about**, because a sentence that simply BEGINS with a name looks identical: "For Chicago Bears news, I would like..." has a capitalized opening followed by normal text too. Lickety cannot tell those apart from the words alone, so it sometimes lowercases a name.

TWO THINGS LIMIT IT. Neither switch looks at anything longer than 10 words, and a term in your Vocabulary is never touched — so adding a name you say often is the reliable fix. If it still catches names you use, turn the second switch off; you will occasionally see a capitalized opening you have to fix by hand instead, which is the trade.

Turn Tidy up off if you want exactly what the recognizer produced. Structural features — line breaks, lists, paths, spoken punctuation — keep working either way; those aren't style, they're things you asked for.

KEEP THIS MANY DICTATIONS controls the local transcript history — 50,000 by default, 0 for unlimited. Older ones are removed as new ones arrive.

It's a plain local database on your machine and nothing leaves it. If you'd rather keep less, lower the number; the trim happens on the next dictation.

LOWERING IT ASKS FIRST. If the new number is below the dictations you already have, Lickety tells you how many of the oldest will be deleted and asks before changing anything — the trim cannot be undone. If you press OK, a second box asks you to type YES before anything is deleted. Cancel at either step leaves the number as it was. AND WHEN YOU REACH THE LIMIT BY DICTATING, the first time a dictation takes you past it nothing is deleted: a notice in the corner says so and stays until you close it. From the next dictation on, the oldest one goes each time — raise the number to keep more. You can type the number with or without its comma: 50,000 and 50000 are the same. Your notes and your statistics are never trimmed.

HOW MANY YOU HAVE is shown right under the box — "Dictations stored now". That is the number the limit counts. The Dictations figure on Home and Stats can be higher: it counts every dictation you have ever made, including ones you deleted from Transcripts.

Backups can optionally include this history, and by default they don't — the archive would contain every word you've ever dictated, and backup folders are often cloud-synced.

OFFER THE TEXT BACK IF AN INSERT MAY HAVE FAILED — on by default.

Some applications accept text and then quietly don't show it. When Lickety can't confirm the insert landed, it shows a small panel with your text and a COPY button, so the words aren't lost while you work out what happened.

It's also a mouse-only path to your last dictation, for anywhere the keyboard is not an option.

CLEAR STRAY MENU FOCUS (ESCAPE) BEFORE INSERTING is a per-app workaround for windows that leave a menu armed — Notepad is listed by default.

LICKETY IS FREE FOR YOUR FIRST 25,000 WORDS, with no limit of any kind and nothing said about one. With heavy use that is about a week; with light use, much longer.

AT 25,000 WORDS YOU ARE TOLD ONCE — a small window explains the paid version and offers the store page. If you'd rather not buy, close it: Lickety keeps working for free at 2,500 WORDS A WEEK, for as long as you like. The week starts again every Monday at midnight. A dictation that crosses the line finishes; the next one is refused with a line on the status bar and in a tray balloon saying when the week starts again. Nothing you have set up is touched: your replacements, lists, notes, history and settings all stay exactly as they were.

The Home screen shows the week's count once the weekly limit applies, and the LICENSE section of the About screen always says where you stand.

THE PAID VERSION REMOVES THE WEEKLY LIMIT. It is one purchase, not a subscription; the price is on the store page linked from About. Your purchase e-mail carries a key — five groups of letters and digits separated by dashes. Paste it into the box under LICENSE and press APPLY; the limit is gone from the next dictation. REMOVE takes it off again.

WHAT LICKETY DOES AND DOES NOT CHECK — said plainly. It checks that a pasted key has the shape of a key, and that is all: there is no activation server, no account, and nothing is sent anywhere, which is the same promise the rest of the program makes. The word count behind the free plan is a file on this PC. In other words the plan runs on trust — the key is your receipt, and we'd rather ask you to be fair than build a program that phones home to check.

WHAT COUNTS: words you spoke that Lickety typed — the same words the Home screen counts. Transcribing an audio file on the Transcribe screen does not count.

Some controls were added while Lickety was being tested and are off the screen unless you ask for them, so the everyday screens stay clear.

SHOW DIAGNOSTIC CONTROLS — off by default, on the About screen under Export logs. On, it shows: the Great transcription button and the rated count on Home; the All / Great / Issues / Unrated buttons and the Great, Issue and Open audio buttons on Transcripts; Accuracy and Reset ratings on Stats; the Great transcription hotkey on Settings; the name pieces under the file name on Transcribe; and four diagnostic switches on About. Their Help topics appear with them. Support may ask you to turn it on while looking into a problem. Turning it off also turns the four diagnostic switches off, so nothing keeps logging or recording out of sight. Report an issue, Export logs and the right-click "Report an issue with this one" are always there — they are how you reach us.

SHOW WHEN REPLACEMENTS ARE USED IN TRANSCRIPTS — on by default, and not a diagnostic. Puts a small line under any dictation on the Transcripts screen that a replacement rule changed, naming the rule, so a rule that fired when it shouldn't is easy to find. Turn it off if the extra line is clutter.