DevThrottleYour dictation keeps getting the same word wrong. Now it tells you which.
Speech-to-text is very good at English and has never heard of your project. Your product name, your tools, your colleague's surname: those are the words that come back wrong, and they are the same words every time. You fix one by hand, send the prompt, and meet the same mistake tomorrow.
A dictionary solves this, but it has a problem of its own. You have to notice the mistakes and remember to write them down, and you are busy.
The list writes its own first draft
So DevThrottle does the noticing. Your Gateway scans your recent dictations for distinctive words the speech model keeps getting wrong, screens out ordinary English, and offers what is left at the top of the Dictionary page. The scan runs nightly, and Scan now runs one on the spot.
You find it under Settings, in the menu behind your name at the bottom of the Cockpit's left rail. While terms are waiting, Settings carries a red count, so you can see there is something to look at without opening the page.
What a suggestion shows you
Each one is a row with the evidence on it:
- The term, as it should be written.
- The wrong spellings it was heard as, from your own dictations.
- A count of how many times out of how many it came out wrong.
Rows arrive ticked. Untick any you do not want this round and press Add, and the term and its wrong spellings go into your lists. Nothing changes until you press it. Dismiss stops a term being suggested again, and dismissed terms can be restored later.
If the screening step cannot be reached when a scan runs, the page says so and holds back the new candidates rather than showing you unscreened guesses.
Why this is safe to accept
Suggestions feed the same two lists you can edit by hand: Vocabulary holds your spellings, and Common mistranscriptions pairs the correct term with the wrong spellings you have seen. The second list does the work, and what it does is narrow. Every change is a find-and-replace of a wrong spelling on your list. A word that merely resembles one of your terms is left exactly as you said it. A transcript with none of your listed spellings in it comes back untouched.
That matters because your dictation should type what you said, and nothing else. An agent acts on the words. A correction you approved is your own edit. A guess at what you meant would be an instruction you never gave.
Teach it once
The dictionary lives on your Gateway, so a term you add covers every voice surface: dictation and Speak on every Director, and voice-note transcription. There is nothing to restart or copy between machines, and a saved edit applies to the next transcription that reads it.
The next time you see a red count on Settings, give it thirty seconds. How the lists and the suggestions work is on Your dictation dictionary, and the dictionary tutorial walks through it.