Systran

/

faster-whisper-large-v3

like 140

Is faster-whisper-large-v3 performs better than large v2 or not?

I tried the large-v2 and the large-v3 model on the benchmark dataset, but found that large-v3 model performs worse than large-v2 model. What I did is that I upgrade the faster-whisper from version 0.9 to version 0.10 and load large-v3 model for transcription task. Is there other configuration steps needed? Thanks