A WIP fabric mod that transcribes audio data from simple voice chat to text using open source speech to text whisper. [https://github.com/GiviMAD/whisper-jni](https://github.com/GiviMAD/whisper-jni)
This mod is entirely client sided. All transcription is done on your computer, so may be limited on low performance machines.
subtitles rendered in top left of screen
No setup should be required, the whisper library is included, and the mod will attempt to download whisper's base model from official sources into your .minecraft instance folder. (Models can be found here: [https://ggml.ggerganov.com/](https://ggml.ggerganov.com/)). If you run into issues follow instructions in github to download the models directly.
The accuracy is medium quality, which is primarily the fault of me processing the audio poorly. If I can figure out how to properly process the audio (help welcome) improving accuracy I will continue to update and improve this mod. (whisper has language translation capability which I would love to support)