Repository navigation
Add background prompt settings to support contextual local ASR models #2509
Eric-Terminal
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
When I use Handy with Qwen3-ASR, it outputs Chinese characters instead of Arabic numerals, which makes reading a bit troublesome. Also, for more specialized fields, for example, when speaking Chinese itself and talking about programming or machine learning, English abbreviations and technical terms get mixed in. Then I took a look: Qwen3-ASR's open-source inference interface supports context, and the transcribe.cpp used by Handy also has context prompt capability. I'm wondering, for models that support context capability, could an optional background prompt input box be added?
I've considered the current project's experimental feature: enabling post-processing. But if it's local, then another model has to be newly loaded, and recognition is also slower. And if using an online API, since local voice input is for privacy in the first place, then making online API requests would mean there's no privacy benefit.
All reactions