[nopelynopington]

If this lives up to the demo it's a huge development for anyone looking to do realistic tts without paying to use an API

[kristopolous]

there's quite a number of pretty low overhead models around that do that in realtime these days.

[MarsIronPI]

But how many of them support voice cloning?

(Genuine question; I haven't seen any other than this one.)

[nickthegreek]

microsoft’s vibe voice.

[MarsIronPI]

VibeVoice (according to the repo description) is currently unavailable due to "misuse". But my impression was that it required a significant (>8GB) amount of VRAM? Or that it wasn't suitable for on-device for devices with low specs.

[nickthegreek]

its unavailable from their repo, but was released with an open license and mirrors exist. I'm not sure what the VRAM req are.

[MarsIronPI]

According to this issue[0] the 1.5B model needs 6GB of VRAM. Meanwhile it looks like NeuTTS is designed to be able to run on CPU, which is nice for older/lower-spec hardware.

0: https://github.com/microsoft/VibeVoice/issues/26#issuecommen...

[foofoo12]

no