What if I'm already self hosting a STT server

[thiswillbeyourgithub]

Hi,

I just discovered your project and I have to say I'm very interested but I am already running a Docker compose powered faster whisper server and was wondering how I could integrate your project with my already on API instead of creating a new one and using double the VRAM.

Thanks!

[KoljaB]

Depends a lot on the project details. It's not possible to share the RealtimeSTT main transcription model since it runs in a separate process.

For integration with external audio processing, RealtimeSTT offers the following options:

• Input:
  To integrate RealtimeSTT with another audio chunk provider, you can use the feed_audio method and set the use_microphone parameter of AudioToTextRecorder to False. For example you could record audio from a browser, send the chunks to your backend, and feed them into RealtimeSTT for transcription.

• Output:
  If you want to process the audio chunks used for transcription, you can access them via recorder.last_transcription_bytes. This can be used to calculate the length of the last audio chunk, determine the volume etc.