collabora / WhisperLive

A nearly-live implementation of OpenAI's Whisper.
MIT License
2.01k stars 274 forks source link

Failed to Transcribe audio chunk #275

Open coding-with-umar opened 1 month ago

coding-with-umar commented 1 month ago

Hi, I’m working on transcribing a live YouTube video. I’m fetching the audio from the live stream and passing it into Faster Whisper. The issue I’m encountering is that after a while, or sometimes even right from the start, it fails to transcribe certain audio chunks, leading to gaps in the transcription. Could you provide some insight into what might be causing this and how I can resolve it?

Here’s the error I’m seeing on the server side: INFO:faster_whisper:Processing audio with duration 00:04.492 INFO:faster_whisper:Processing audio with duration 00:14.476 ERROR:root:[ERROR]: Failed to transcribe audio chunk: cannot access local variable 'last_segment' where it is not associated with a value

Note: I’ve tried every model size, and I’m working with Urdu language transcription.

makaveli10 commented 1 month ago

Would you mind sharing the youtube link so as to reproduce the issue on our side?

coding-with-umar commented 1 month ago

Would you mind sharing the youtube link so as to reproduce the issue on our side?

sure. Here is the link to that YouTube live video: https://www.youtube.com/watch?v=O3DPVlynUM0

Its News channel.