Skip to content

Android: Default to a larger voice typing model #12352

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: dev
Choose a base branch
from

Conversation

personalizedrefrigerator
Copy link
Collaborator

@personalizedrefrigerator personalizedrefrigerator commented May 27, 2025

Summary

This pull request changes the default voice typing model from whisper-base-q8_0 to whisper-small-q8_0.

The small-q8_0 model is larger than base-q8_0 and has better accuracy. This is particularly important for non-English languages (e.g. French) that have low accuracy with the base model. However, transcription with this model is slower. See the model release page for more details.

Testing

Android 13:

  1. Open a note.
  2. Start voice typing.
  3. Verify that the dialog shows "Downloading English language files..."
  4. Verify that after the model finishes downloading it's possible to start voice typing.

@personalizedrefrigerator
Copy link
Collaborator Author

personalizedrefrigerator commented May 27, 2025

Converting to a draft — with English-language input, the small-q8_0 model has been observed to occasionally insert "It's a note taking application." when it should insert nothing. This may be due to the prompt included with the model.

Update: This seems to be roughly just as much of an issue as with the base-q8_0 model, so is not a regression.

@personalizedrefrigerator personalizedrefrigerator marked this pull request as ready for review June 2, 2025 12:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant