Heya, I just made this because finally AI audio is good enough to actually provide useful listening practice for languages where material is incredibly scarce. Just Cantonese for now, but hoping to expand the model to others too. Feedback welcome!

Mozilla Common Voice has a dataset with hundreds of hours of recorded Cantonese sentences: https://mozilladatacollective.com/datasets/cmu5x3i5w00d6o107... Hardly scarce enough to require AI audio for listening practice.