Finetuning OpenAI's Whisper models on a big dataset for Persian speech recognition. This dataset is around 1,750 Hours of Persian speech data with different accents and speaking styles.
We are training several models by mixing several existing datasets and two big datasets we collected for Persian Automatic speech recognition. This is a personal place to store some codes until the final version of the code and models. But initial results are awesome!
All the code, trained models and datasets will be shared freely.
Loading...