Every tool that touches your words
Where does your
text actually go?
Most dictation tools record you, send the audio to a company's servers, transcribe it there, and send the words back. Lickety does the middle part on your own processor. There is no round trip because there is nowhere to go.
The dotted line is the edge of your computer. On the left, the words cross it. On the right, they don't.
What "local" means here, precisely
The speech model is a file that ships with the program and sits on your drive. When you release the hotkey, the audio goes to that file, on your CPU, and comes back as text a second or two later. It is the same thing that would happen if you unplugged your network cable — because that's a test you can run.
Pull the ethernet, turn off Wi-Fi, and dictate. It works exactly the same. That's the whole claim, and it takes about ten seconds to check.
- No account. There is no sign-up screen, because there is no server holding a list of who you are.
- No upload. Audio is transcribed and discarded. It isn't stored, sent, or used to train anything.
- No screen capture. Some cloud tools screenshot your active window every few seconds for "context." This one has no reason to look at your screen and doesn't.
- Your history is yours. Transcripts are kept in a database file on your PC so you can search them. Delete the file and they're gone.
- Your word lists are files. Plain JSON, with import and export. Not a row in a database you can't reach.
When the words themselves are the problem
For a lot of people local-versus-cloud is a preference. For some it is the whole question, and it is settled before any feature gets compared: a lawyer dictating a client matter, an accountant working through somebody’s return, a clinician, an HR manager writing up a grievance, anyone under an NDA. The content of the sentence is the confidential thing, and a tool that sends sentences to a server is a tool that has to be cleared first.
Lickety never raises that question, because there is nothing to clear. The audio is turned into text by a model running on your own processor. There is no account to create, no server to send anything to, and no setting anywhere that turns uploading on. Unplug the network and it works exactly the same — which is also the easiest way to prove the claim rather than take our word for it. Your dictation history and your notes are encrypted on your own disk, dictations are kept out of Windows clipboard history and the cloud clipboard, and the history can be locked behind a password.
What we are not doing is claiming a certification. No audit badge, no compliance programme, no promises about your obligations — those are yours and your firm’s to judge. What we can tell you is the mechanism, which is the thing an IT department or a managing partner actually asks about: it runs here, and nothing goes out. Ask the same question of any cloud tool and read the answer carefully.
The uptime argument nobody makes
Privacy is the obvious reason to want local, and it's the one every competitor puts on their homepage. The less obvious reason is better: a cloud dictation tool has a failure mode where everybody's dictation stops at the same time. One capacity incident on their side and the tool you were mid-sentence with is simply gone.
A tool with no server can be slow, and it can have bugs, and both of those are your problem to report. But it cannot be down, throttled, rate-limited, degraded during a traffic spike, or discontinued because a company changed direction.
Related
It learns your words
Seed your customers', vendors' and coworkers' names before it listens instead of repairing the output afterwards.
Read →The five-year maths
$12 a month billed annually is $720 over five years. This is $99, once, and then nothing happens ever again.
Read →Dictate in a remote session
Full-screen RDP is where most dictation tools quietly stop working. Here it was the first thing that had to work.
Read →Lickety — $79, once
Windows 10 & 11. No account, no upload, nothing renews.