Best you can get with the audio quality. The mic wasn't good either. Lowkey just hard to make a good model out of audio thats from an old ass anime.
Dataset: 13:43
Sample Rate: 32k
Pretrain: Og Pretrain
Batch size: 4
Made on Kaggle
Credit:
If you use this model please credit me on YouTube: @Th4t-Ai-Guy