I’ve tried a number of times to import Cantonese audio with automatic transcript and it is not working for me. Has it worked for anyone else? It takes a very long time and then get an empty lesson. I’ve had really good results using Whisper on Google Colab on the standard medium model though, as long as I give it an --initial_prompt “你哋嗰度嘅天氣點呀?” because I believe usually the first 30 seconds of audio isn’t enough for it to autodetect the correct form of Chinese most of the time…
Also, Is there a way to upload a Whisper generated transcription with time stamps and have it match uploaded audio without having to go through in sentence mode and re-match the audio to transcript?
Make sure to have the audio file and the matching subtitle file in place before clicking “Save and generate lesson”. That should preserve audio sync.
We will investigate the issue, thanks for reporting.