Slow AF

[agesagem]

Dear Zonos creators,

just tested Zonos today and it seems it is around 10 to 100x slower than tts_models/en/ljspeech/vits.
-- same hardware -- RTX 4060ti
The good thing is it picks up from a sample wav voice given, but the slowness makes this advantage completely irrelevant.
Am I doing something wrong that it is so slow - or can this be speed up?

[DarkAlchy]

I'm on Windows and can no longer even run this thing and core zonos demands Python 3.12 which then breaks ComfyUI and torch compile. Between it, and this, it won until someone figures out why torch compile only works on 3.11 and torch 2.6 in ComfyUI. Someone I gave the info too said they were looking into it as there are a bunch of us out here.

Not sure why this had to force (core zonos) 3.12 on us.