guides
vocabulary
Teach the transcriber your names and jargon — and how it should recover from mishearing them.
Speech models are great at everyday words and mediocre at your words — product names, company jargon, colleagues' names. Settings → vocabulary is where you fix that.
Each entry has two parts:
- a term — the correct spelling: SpaceX, fnatic, Kubernetes…
- optionally, sounds like — how the transcriber tends to mishear it:
space ex; space x. Separate alternatives with
;or,.
The term alone already biases transcription toward the right spelling. Adding "sounds like" goes further: after transcription, known mishearings are corrected to the term before the transcript is saved.
Names from your speakers directory are included automatically — naming a colleague also teaches the transcriber to spell them properly.