ggerganov
Streaming to a buffer or queue
Is there a way to use stream.cpp but stream the output into a buffer or queue. I want to implement a real time speech to text but use the transcription for something in real time. How would I be able to do this? Thank you so much for the help!
The short answer is there is no functionality built in to do this. The streaming example is just calling whisper_full and whisper_full_get_segment_text repeatedly and printing the result of the latter. If you want to "stream" the output to a buffer etc., you need to do something different with the result of whisper_full_get_segment_text than print it – for that you need to use your C++ skills. As far as the required domain knowledge goes, you can only really "stream" in real time as the result is being decoded via a greedy search, not the default beam search.
For python, see https://github.com/davabase/whisper_real_time.
With the library speech_recognition, it's quite easy.
For C++, I think ffmpeg api should be the best way.
Maybe it's easier to re-implemnt speech_recognition in C++.
