Qwen3-Omni ASR with Transformers only transcribes first 30s of long audio

#21
by AndreosLXIX - opened

I tried the Speech Recognition cookbook but ran it with Transformers instead of vLLM.

On short audio files it works fine, but when I tested with a longer file (~84 seconds), the transcription stopped after about 30 seconds.

What could be the reason for this?

I meet the same problem with vLLM , speech recognition works for first 30 seconds only

Seems to be a transformers configuration issue, which has been solved recently
https://github.com/QwenLM/Qwen3-Omni/discussions/36

Seems to be a transformers configuration issue, which has been solved recently
https://github.com/QwenLM/Qwen3-Omni/discussions/36

thanks very much

Sign up or log in to comment