You need to train an unsupervised voice classification model on the fly before sending it to Whisper.

→ View original post on X — @dfintelligence