Vocabulary and Advanced Settings

This page keeps a few tuning-oriented capabilities in one place: Hotwords, Text Replacement, Speaker Profiles, and the Advanced Settings under LLM Polish. They are not first-run requirements, but they become useful when you want more control over recognition, speaker matching, or cleanup behavior.

Best for

• Terms, names, or product words that are repeatedly transcribed the wrong way
• Workflows where you want consistent wording after transcription
• Known speakers you want Sona to suggest or match in future transcripts
• Users who want to tune Auto-Polish or custom polish context beyond the default flow

Open the right settings area first

• Settings > Vocabulary: manage Hotwords, Text Replacement, and Speaker Profiles
• LLM Polish > Advanced Settings: manage Auto-Polish, frequency, keywords, scenario presets, and custom context
• If you still need the basic polish and translation flow first, return to AI Polish and Translate

What Hotwords are good for

• Add repeated terms to Hotwords when recognition keeps drifting on the same names or phrases.
• Hotwords are entered one per line, and weighted entries such as Term :2.0 are supported.
• Right now this capability is especially relevant for Transducer and Qwen3 ASR models.

What Text Replacement is good for

• Use Text Replacement when the same term keeps appearing in inconsistent forms after transcription.
• It works best for repeated cleanup patterns, not as a replacement for reviewing segments one by one.
• If your focus has shifted to editing transcript segments directly, continue to Edit and Playback.

What Speaker Profiles are good for

• Create a profile for a known speaker, then import one or more local reference audio samples for that person.
• Sona normalizes imported samples into app-managed audio and uses the profile for candidate suggestions or automatic matching when speaker models are configured.
• A profile can show as ready for automatic matching, limited to suggestions, or not ready yet depending on usable sample count and duration.
• Projects can choose which speaker profiles are active, so a client, course, or meeting series can keep its own speaker set.
• Speaker profiles are not a promise of perfect automatic attribution. You can still confirm suggestions, assign another profile, or reset a group to anonymous from Edit and Playback.

When to open Advanced Settings

• When you need more than a one-off LLM Polish action and want to manage Auto-Polish.
• When you want to tune Auto-Polish Frequency, keywords, scenario presets, or custom context.
• When provider and model setup is already done, but you want polish behavior to match a more specific use case.

Useful notes

• Advanced Settings still belongs to the AI Polish and Translate workflow. It does not replace the base provider and model setup.
• Speaker profile readiness affects candidate quality and automatic matching confidence, not whether you can manually assign a profile.
• If you mainly want a transcript-side recap rather than polish behavior, switch to AI Summary.
• If a feature is already configured but still behaves incorrectly, continue to FAQ and Troubleshooting.