vosk-api
vosk-api copied to clipboard
Demo for fast recognition of a big file
I'm not sure if I'm doing anything wrong. I wanna transcribe an audio file that's one hour long. But it seems like vosk is transcribing in realtime and the whole file is gonna take an hour. Is there a way to do this faster? Should I maybe split up the audio file in several parts and do those concurrently?
Usually it is two times than realtime unless your hardware is slow.
To process fast you can split audio on chunks with small model and then recognize accurately with a big model in parallel.
@vadimdddd please implement a demo for that.