Very early checkpoint, but it's a good model/finetune.
Base model: MiMo Audio
Prompt format:
Emotion: <emotion>
Text: <text>
Example:
Emotion: intense anger, rage, fury, hatred, and annoyance, speaking without any accent
Text: You know what? I'm done. I'm done with your excuses. (sharp exhale) Every single time, it's the same, and I actually believed you'd change. (voice cracks slightly) God, I'm such an idiot for trusting you again.
Training code (private):
https://github.com/fakerybakery/mimo2
Voice cloning is not supported yet