Great Scott Labs

Comparison

Where Lickety wins,
and where it doesn't.

Most dictation tools handle ordinary speech about the same. On plain English, a big cloud tool may match this one or beat it. So the honest question isn't which is most accurate — it's which one gets the hard words right. And there are two kinds of hard word.


Your words: Lickety wins

Your customers, your vendors, your coworkers, your internal systems. No dictation tool has seen these names, because they aren't in anyone's training data. When you say one, the tool is guessing at sounds.

The $29 local tool wrote "Fastenal" three different ways in three attempts. The $144-a-year cloud tool wrote "Claude" as "Clawd," because nobody had told it the word. Neither can be taught, because their dictionaries need a stable wrong answer to catch.

Famous words: Lickety loses

Here's the part we'd rather tell you than have you find out. Unusual words that are well known enough to be in a big model's training — "Snuffleupagus" is the test we used — the cloud tool gets right every time. Lickety gets it wrong, and differently each time.

The reason is simple: a large cloud model has seen the word. A smaller local model hasn't fully learned it, so it re-guesses the sounds. On a well-known hard word, the big cloud tool is better. No spin.

The one thing Lickety can do that they can't is be taught the word it loses on. Add it once and it's right from then on. The cloud tool gets famous words right on its own — but the famous words it happens to get wrong, you can't fix.

The honest picture

  Your words
customers, vendors
Famous hard words Ordinary English
Lickety
$99 once, local
Wins. You teach it the word before it transcribes. Loses on its own — but you can teach it. About the same.
Cloud tool
$144/yr
Loses. Can't be taught your words. Wins. The big model already knows them. Strong.
Local competitor
$29 once
Loses. A dictionary can't catch a wrong answer that keeps changing. Consistent, but wrong. Fine.

Tested by hand, one voice, one machine. The method is on the vocabulary page so you can repeat it.

Try it yourself

Say your biggest customer's name three times into whatever you use now, and count how many spellings you get. That one test tells you more than any review, including this one.

Page note: the draft in the project uses "$144-a-year cloud tool" and "$29 local tool" rather than brand names — lower risk, and it keeps the reader focused on their own test. Using real names raises credibility and catches the "[competitor] alternative" searches, which is most of this page's traffic. Worth a minute of the attorney's time while the naming question is already open. The filename here assumes the named version. Delete this box before publishing.