[REQUEST]: A way to select audio input channel for users with an audio interface

[enhancedrouting]

In a few words, describe your idea

Be able to select the input channel w-okada uses with audio interfaces

More information

Spent about two hours trying to figure this out after the new version of w-okada had no sound in with the setup that had worked before. My best guess is that w-okada used to just take two channel audio interfaces as stereo in which meant the mic worked regardless of which channel the mic was connected to. Now it doesn't seem to do that anymore. I eventually figured out my mic needed to be plugged into the first channel on my audio interface and then it worked. If w-okada is going to be channel specific in that way it would be helpful to be able to select the channel it pulls audio input from. Alternatively, if this change isn't already in the documentation (in which case I missed it) adding a sentence to let users know that, if they are using an audio interface, they need to be on channel one, would save some troubleshooting time. Thanks.

[Kuuko-fokkusugaru]

RVC works perfectly fine with most kind of inputs and outputs. If your input doesn't appears in the software but it does on windows then there may be an issue with them. A screenshot of the inputs and the one missing would be helpful.
One of the things that prevented from showing inputs properly was nvidia broadcast, which doesn't seems to be fully compatible with RVC for some reason.
You also omitted adding which version are you using since there is actually no new one wince months.

[enhancedrouting]

I'm not on Windows, I'm in Linux: Debian Bullseye.

If it's NVIDIA broadcast/RVC doing this just adding a note about a possible issue with audio interfaces only accepting channel 1 (no idea if that's linux specific, no win system to test against) in the docs would work, just be better if the next person doesn't have to troubleshoot this like I did.

I didn't mention a version because I can't find any version information for w-okada/voice-changer either in the cloned git folder or in the web interface. I didn't install by downloading a .zip file so I wasn't provide that version number, I am running whatever version the current git folder is. When I say it worked previously I mean about 7 months ago when I last did a clean install from a git clone. I also didn't mention a version because I'm not asking for help troubleshooting this now that I see what voice changer appears to be doing. I worked around it by changing which channel my xlr mic was plugged into. It would be nice to have a channel selection option in the future or, if that's too much work for too few users with audio interfaces, just a note that behavior may be odd with multi-channel audio interfaces and users should try channel 1 if their microphone doesn't work.

Currently training a model at near the max of my GPUs capacity. Don't want to crash training. Once it's done I'll relaunch the voice changer and post a picture of what I see.

Not trying to talk down to you in anyway, but to make sure we're on the same page, for high quality audio input some people (including me) use audio interfaces which look like this: https://us.focusrite.com/products/scarlett-2i2-3rd-gen. The two circular XLR plugs on the front are inputs for studio microphones. In the past voice changer seems to have taken both inputs, probably by treating the two channels as a single input as many programs do, which might explain why it worked without the ability to select between the two input channels. Now it doesn't do that anymore. Instead it ignores the second channel and only passes in the first channel. This behavior is unique to voice changer on this system. Everything else either exhibits voice changer's old behavior of treating both channels as a single input or, in the case of audio specific programs (Audacity and Reaper) allows me to select between channels if I want to.

In voice changer, only the audio interface can be selected, per channel selection is not an option (and never has been, hence my suggestion) so it's impossible to select the second channel. That is perfectly fine if that's too much work to fix, but it would be helpful if this was noted in the documentation so people using these devices don't have to discover the fact that only the first XLR input may work on their audio device themselves.

[Kuuko-fokkusugaru]

I guess there may not be that many people using these types of devices because, after all, RVC doesn't even make much sense other than having some control over the volume and mic activity. The quality of the input audio pretty much have zero effect in RVC because this is a voice converter and not a morpher. So whatever input audio comes in, as far as it's "understandable" by RVC, it will be converted into a new thing. In fact, it's often better to use mid-range mics instead of higher-end ones because they pick way fewer small sounds around due to having less precision and quality. This makes it easier for RVC to do its job.
That said, obviously, you are free to use it whichever way you want to connect to RVC. Alternatively, if you are into professional audio, you can easily redirect the right channel into a specific virtual cable and use that as RVC input.
Since you don't know which version of RVC are you using, it's hard to help further. The reason why I asked is because there are two of them, RVC 1 and RVC 2. For most people, RVC 1 works better with higher quality conversion and fewer issues, which surely may be the version that you used months ago.

[Jeffrey-deng]

RVC 1 and RVC 2

RVC 1 and RVC 2
means voice-changer v1 and voice-changer v2
or means rvc model v1 and rvc model v2 ?

thank u

[Kuuko-fokkusugaru]

RVC v1 and v2 are the different versions of the software. Isn't related with the voice models version. Both works with the same models just v2 is newer and supposed to use less resources (in my case uses the same and have slightly less quality). For some people v2 sounds better with their models, for some others v1 does.