Hacker News

A 4.5M-parameter TTS model runs 70x real-time on CPU and cost just $10 to train

43 points by mrkn1 ago | 4 comments

ilyaizen |next [-]

Hi there.

I made CopySpeech, a desktop app that reads copied text aloud. Copy the same text twice to hear it.

The default voice is tiny-voca, a 4.5M-parameter model. In my tests, it starts speaking about 0.1 seconds after a warm start. I'm happy with how fast it is.

Repo: https://github.com/ilyaizen/CopySpeech Site: https://copyspeech.ilyaizen.link/

ig-88ms |previous [-]

I was irritated by the childlike voice at first, but they acknowledge that they were aiming for a twenty something female voice and plan to build a more mature voice.

chandradot99 |root |parent |next [-]

yeah pretty much same for me, here are other small models which you can try kitten tts(about 7m), kokoro(82m but still runs on cpu).

mrkn1 |root |parent |previous [-]

i dont like it either, but you can train your own voice with 10h of audio.