Skip to content

v1.0.2

Choose a tag to compare

View all tags
@Matthew-Hughes3488 Matthew-Hughes3488 released this 29 Sep 22:50
· 1 commit to main since this release
v1.0.2
6bf8697

Patch Release Notes: Bundled Whisper Models

Download

Download the Audio_Transcriber.zip near the bottom of the page
Inside is a guide on how to use it :)

Overview

This patch release introduces a significant enhancement to the Audio Transcriber application by bundling Whisper models directly into the app. This ensures seamless transcription functionality without requiring runtime downloads, making the application more reliable and user-friendly, especially in offline environments.

Key Features

  • Offline Functionality: Whisper models are now included in the application package, eliminating the need for internet access during transcription.
  • Simplified Setup: End users no longer need to manually download or configure Whisper models.
  • Improved Reliability: Bundling ensures that the correct model versions are always used, reducing potential compatibility issues.

Technical Details

  • Whisper models are stored in the models/ directory within the application package.
  • The application dynamically loads models from the bundled directory, ensuring compatibility with the directory-based build method.
  • The ffmpeg binary and Whisper assets are also included in the build for a complete offline experience.

Build Instructions

To create a build with bundled Whisper models, use the following PyInstaller command:

pyinstaller --name "Audio Transcriber" app.py \
  --add-data "models:models" \
  --add-binary "/opt/homebrew/bin/ffmpeg:." \
  --add-data "/Users/matthewhughes/development/AudioTranscriber/venv_audio/lib/python3.12/site-packages/whisper/assets:whisper/assets"

Usage Notes

  1. Extract the dist/Audio Transcriber/ directory from the build output.
  2. Navigate to the extracted directory.
  3. Run the application: 
    ./Audio\ Transcriber
  4. Ensure all required files (e.g., models/, ffmpeg, and Whisper assets) are present in the directory.

Benefits for End Users

  • Ease of Use: No additional setup steps are required; the application works out of the box.
  • Offline Capability: Users can transcribe audio files without an internet connection.
  • Consistency: The bundled models ensure that all users have the same transcription experience.

Known Limitations

  • File Size: The application package is larger due to the inclusion of Whisper models and dependencies (1-2GB).
  • First Run Performance: The first run may take slightly longer as the application initializes the bundled models.
  • Model Selection: Only includes tiny, base, and small models to keep package size manageable for distribution.

Conclusion

This patch release marks a significant step forward in making the Audio Transcriber application more accessible and reliable for all users. By bundling Whisper models, we have eliminated a major dependency on internet connectivity, ensuring a seamless transcription experience in any environment.

Assets 3
Loading