[P] TorToiSe - a true zero-shot multi-voice TTS engine

I'd like to show off a TTS system I have been working on for the past year. I've open-sourced all the code and the trained model weights:
https://github.com/neonbjb/tortoise-tts

This was born out of a desire to reproduce the original DALLE with speech. It is "zero-shot" because you feed the text and examples of a voice to mimic as prompts to an autoregressive LLM. I think the results are fantastic. Here are some samples:
https://nonint.com/static/tortoise_v2_examples.html

Here is a colab in which …

machinelearning tts voice