My voice changer keeps lagging and is super delayed when using models

[random ermine]

Pls any help would be appreciated

[hollow folio]

Is this happening on downloaded models specifically? And what is your GPU?

[random ermine]

Yea only downloaded models, I asked someone for help and they said I downloaded the wrong version for my gpu but with the right one its still lagging

I have a 6700xt

[hollow folio]

You need to export downloaded models to onnx

[random ermine]

How do you do that

[hollow folio]

https://rentry.co/VoiceChangerGuide#uploading-models-amdintel-gpus-and-fixing-its-laggy-issue

This should help, lmk if anything undesired happens.

[random ermine]

Ok it fixed it, one more thing though is there a way to lower delay or is it because of my settings im getting like a 300ms delay

these are my current settings rn*

[hollow folio]

Make sure you're using RVMPE_ONNX. After that idk what else could be done - AMD cards have more delay than NVIDIA cards in W-Okada even with ONNX conversion

[random ermine]

What is that?

[hushed bronze]

Ayo? @random ermine level 2 !!!

[hollow folio]

You're currently using it. It's the pitch extraction method, labeled as "f0 det:"

[random ermine]

Oh ok, is there anything wrong with my chunks or extra or anything that could cause a big delay?

[hollow folio]

Maybe reduce extra to 8096? That could help with delay a bit.

[random ermine]

oh yea that helped thank you. Also one last thing do you know how I could export the voice as a microphone?

also to reduce background noises while im typing etc

[hollow folio]

Oh you need a Virtual Audio Cable

The guide links you on the main virtual audio cable download pages

[hollow folio]

Activate Sup2 and increase S Thresh a bit

[random ermine]

Alright thats all my questions appreciate the help!

[hollow folio]

You're welcome :)

Make sure to mark this post as solved too

[vivid granite]

isnt client mode worse than server?

[hollow folio]

No

Because it has sup2