About
This plugin allows you in conjuction with a general-purpose speech recognition model to transcribe your voice messages to text.
In order to make use of this plugin, you need to have at least one of the following models installed:
OpenAI Whisper
- Website: https://github.com/openai/whisper
- Installable by:
pip install -U openai-whisper
Faster Whisper
- Website: https://github.com/SYSTRAN/faster-whisper
- Installable by:
pip install -U faster-whisper
Additionally you have to checkout the following Gajim branch: https://dev.gajim.org/mesonium/gajim/-/tree/stt_voice_messages
Hint
The plugin is very much POC at this stage!
Currently a chosen model will be on first downloaded in the background, during which Gajim's UI may not respond.
Typical model sizes are in case of OpenAI Whisper:
| Multi Langual Model | Download Size |
|---|---|
| Tiny | 70 MB |
| Base | 140 MB |
| Small | 460 MB |
| Medium | 1.4 GB |
| Large | 2.9 GB |
TODO
- Offer multiple models
- Add various model settings
- Model receiving
- Hint model download state
- Allow to change model download location
- Allow to use local models
- Database Handling
- Store transcribed messages in a DB
- Option to delete DB
- Update UI
- Make it more pretty
- Show progress bar
- Highlight words on playback