Glossaries teach Tota the names, jargon, and acronyms you actually use, so "Vercel" doesn't come out as "versel". A glossary works in two layers:
- Live biasing — the active glossary's terms are fed to the Whisper model as a hint while it transcribes, nudging it toward your vocabulary.
- Phonetic matching — after transcription, terms that were still misheard are fixed by sound-alike matching (e.g. "and tropic" → "Anthropic"). Toggle this in Settings > Corrections > Phonetic vocabulary matching.
Using Glossaries
- Go to the Glossaries tab in the sidebar.
- Pick a glossary to make it active, or create your own and add terms.
- Keep glossaries focused: the live hint has a small budget (roughly 70–90 words), and a meter shows how much of it your glossary uses.
No glossary is active until you pick one — biasing stays off by default, and "no glossary" is remembered across launches.
Built-in Starter Glossaries
- Development — common programming tools and product names.
- Names — a starting point for people's names.
- Medical — clinical terminology.
All built-ins are fully editable and deletable.
Auto-Pick by App
Optionally, Tota can switch glossaries based on the app you're dictating into — for example Slack → Names, VS Code → Development. Turn on Auto-pick glossary by app in Settings > Corrections (off by default). Apps without a match use your selected glossary.
Glossaries vs. Dictionary
The Dictionary learns corrections from your edits automatically; Glossaries are terms you provide up front. Use a glossary for vocabulary you know Whisper will struggle with, and let the dictionary catch the rest.
