Parallelism STT Factory.

[bfrisco-raft]

Looking for some feedback to see if this makes sense.

I am attempting use the Multi-processor support in RealtimeSTT, and would like to keep a thread open that takes in audio segments using the .feed_audio(), then transcribe and read out on another thread.

First, does this make sense to use it in this way.

Second, if so, what would be the best practices, as I am having difficulty handling many different audio formats to push into RealtimeSTT. Faster_whisper gets around this by using their audio module to convert to binary on the fly, but RealtimeSST feed_audio handled only PCM np.arrays from what I read.