faster-whisper-server
faster-whisper-server copied to clipboard
Make some changes to the project
I think that in src/speaches folder it's time to add separation of functionality by folders for example: stt/ tts/ common/
Please remove the Piper dependency.
Remove the use_batched_mode parameter, always use the default BatchedInferencePipeline(model=whisper) this is faster and gives better results, prevents hallucinations on long files