WER is high：synthesized audio words have a high error rate #178
Description
I use Whisper's base model to calculate the word error rate for the audio synthesized by Vallex, but the word error rate is as high as 0.22, is this normal?
{{ message }}
I use Whisper's base model to calculate the word error rate for the audio synthesized by Vallex, but the word error rate is as high as 0.22, is this normal?