command-line non-interactive possible?

[njarecki]

Issue Type

Feature Request

vc client version number

1.5.3.5

OS

Debian

GPU

A100

Clear setting

yes

Sample model

yes

Input chunk num

yes

Voice Changer type

RVC

Model type

f0

Situation

Hi! We want to run this on a server and take input audio files and convert them at command line without the gui. Is it possible? We have a small budget for this if helpful. We do it currently with so-vits but would love to with rvc ! I poked around but was too dumb to figure out how

[w-okada]

currently not supproted. But in my roadmap.

[njarecki]

Is there someway I can do it? What I notice is that using batch infer of standard distribution there’s a bunch of startup time every time I run it. Whereas with mangio-crepe for example I run and enter infer interactive and it’s over double the speed So I’m thinking there’s some launch overhead tou are overcoming - It goes from 14 seconds to an inference to 7 seconds on a t4 with dio, 10 with mangio crepe for a 30 second file. As mentioned this is meaningful to what I’m working on, so would be happy to pay to make this happen if it would help speed it along In any case great work ! 

[scorp2007]

is it still not available?

[edmundman]

bump

[cvillela]

Would be great!