it's not working well in Sonoma 14.0

[raultaboraz]

Hello,

First, congratulations for this idea? Much better than terminal commands... I have bought it but I have a problem

I have transcribed a 1h m4a file, It took 40 minutes (using Large model) but the SRT result is very poor, only few lines of 1 hour conversation is not correct.

How can I investigate what happens here?

I have tried whisper (openai) and I am able to get a transcription from that file.

Thanks

[raivisdejus]

Most likely something with the model. Things to try:

• Get the latest version 1.1.0
• Try shorter audio to see if model works fine in principle
• Try different model size
• Try different Whisper type (Whisper or Whisper.cpp)
• Try specifying language, whisper models can some times guess language wrong and then you get garbage.