Has Anyone Successfully Fine-Tuned Whisper for a Local Language for better accuracy

Hello everyone,
am fairly new to Al and coding, and Im curious about fine-tuning OpenAl’s Whisper model to improve its accuracy for a local language. Has anyone here successfully fine-tuned Whisper? If so, how did you do it? What tools, frameworks, or techniques did you use? What method work best? tried doing it my self on colab but I coulddnt seem to make it work, to begin with just used common voices from Mozilla to see if it was even possible, maybe it is my limitation, but just wanted to ask if anyone have done it and could guide me a bit:) 'd really appreciate any insights, experiences, or resources.

that could help!
Thanks in advance!

1 Like

I’ve heard that fine-tuning Whisper is difficult, but I don’t know what that really means…

The links provided by John6666 are really good, but if you want to know more, please reach out to me. I have already successfully fine-tuned whisper for a local language.

1 Like