[ISSUE for v2]: Outputted voice echoes

[Aadit420]

Voice Changer Version

vcclient_win_cuda_2.0.61-alpha.zip

Operational System

Windows 11

GPU

Nvidia RTX 3060 Mobile

CUDA Version

cuda_12.6.r12.6/compiler.34841621_0

Read carefully and check the options

• If you use win_cuda_torch_cuda edition, setup cuda? see here
• If you use win_cuda edition, setup cuda and cudnn? see here
• If you use mac edition, client is not launched automatically. Use chrome to open application.?
• I've tried to change the Chunk Size
• I've tried to set the Index to zero
• I've read the tutorial
• I've tried to extract to another folder (or re-extract) the .zip file

Does pre-installed model work?

No

Model Type

RVC

Issue Description

The outputted voice echoes. I've tried setting my sound source to Server mode and change APIs but the application would crash with something like "mismatched sampling rate".
Sadly, I can't send a .wav file here to show you what I mean by the first part, but the ending is as if it's chattering to itself. I'm not sure how to describe it.

This does not happen in the Google colab version, wich is even more strange

Application Screenshot

Logs on console

G:\vcclient_win_cuda_2.0.61-alpha>main.exe cui --https true --no_cui True
checking the modules...

                      vcclient

---

Application | https://192.168.1.19:18000
Log(rich) | https://192.168.1.19:18000/?app_mode=LogViewer
Log(text) | https://192.168.1.19:18000/vcclient.log
API | https://192.168.1.19:18000/docs
License(js) | https://192.168.1.19:18000/licenses-js.json
License(py) | https://192.168.1.19:18000/licenses-py.json

Please press Ctrl+C once to exit vcclient.

[Kuuko-fokkusugaru]

Of course, if you set the input device to be your system, it will listen to every sound that your pc does. So everything being out through monitor will also be listened by the system input and create a loop.
You need to set a microphone in the input field, not your windows system sounds.

[Aadit420]

So there's no way for it to not loop with itself? That's a shame, especially when this issue did not happen with google colab under the exact same conditions. However, you are right, and when I converted the youtube video I was testing with and played it using the file option, no echo was detected.

My guess is that the colab version just funnels the sounds from whatever you select and nothing else, especially when you choose the "tab" option, while no such option exists in the native client.

[Kuuko-fokkusugaru]

The way to not loop it's to not select the system sound or to simply disable monitor. It's there any particular reason for you to select system sound instead of your microphone? If you want to convert audio other than from a microphone, there is a file option which allows you to select an audio file to play and convert.
You can also consider just disabling the monitor setting it to none. The converted audio will still go through the VAC output just fine, you will simply not hear the preview which will avoid getting it again in a loop.
Basically, the way that it's acting up is expected because your configuration is not the right one.