About the new audio transcription AI models: Voxtral-Small & Voxtral-Mini

[despairTK]

This open source model was released in recent days, and I did some audio transcription tests here: https://huggingface.co/spaces/MohamedRashad/Voxtral

I tested Japanese, English, and Korean, and the transcription results were perfect, much higher than whisper. I can even say that the Voxtral open source model is the best choice for transcription at present. Especially in some audios with heavy accents, the correct transcription results can be obtained, such as a Spaniard speaking English not very fluently with a heavy local accent.

I don't know if you will try this open source model. I think it will replace whisper in the future.

[libero08]

Well i think here we have a problem for splitting , anfd formating a srt file.
Have you do some workaround?
UPDATE:
As i thougth ... pipeline are broken...
ASR ...
text are more performat, but for good performace is better now this one!