Dia TTS

DIA is a 1.6B parameter text-to-speech model developed by NARI Labs, integrated and deployed to the hub using the PyTorch Model Hub Mixin. DIA generates highly realistic dialogues directly from transcripts. You can fine-tune the output based on audio, enabling control over emotions and tone. The model is also capable of producing non-verbal communications such as laughter, coughing, throat clearing, and more.

Dia TTS is a text-to-speech (TTS) model developed by Nari Labs, capable of generating ultra-realistic conversations directly from text transcripts, supporting emotion and tone control via audio, and generating non-verbal communication. It also features zero-shot voice cloning.