Hacker News Reader

history prompts are just unconditionally generated TTS from the same model. any of those can be used as history, but for convenience 10 are provided for each language (to generate things with consistent voices)

View on HN

So the history prompts are collections of text/audio pairs?

history is semantic, coarse and fine. so essentially the same thing thats getting generated just using it as an input before the generation

So how do you clone an existing speaker's voice? That's the part I don't get.

Reply on news.ycombinator.com