Show my avatar without vrchat

[Plushie23131]

Hello! I have a question. Can I use FoxyFace with VSeeFace or another similar program to track my face without VRChat or to display my avatar and emotions without VRChat?

[Jeka8833]

There's no such feature at the moment. I've considered this question, but it seemed to me that there are too many VTube programs already, and many of them use MediaPipe's neural networks anyway.

I'm not very familiar with VTube, so I might be mistaken, please correct me:

• I'm not sure if the VMCProtocol supports tongue direction.

I'm open to discussing current VTube programs so I don't end up creating something that already exists.

[Jeka8833]

In theory, I could create a separate text field for each parameter in VMC to avoid the headache of unifying names.

But I'm curious, do VTube programs show what can be passed to the avatar (or does the user need to import the model into Unity or Blender to find out the parameters)?

[HelikeLong]

Sorry, been all over the place here.
In VRM models it's mandatory to use the naming conventions, otherwise it don't work properly.

In Warudo the plataform can translate the expression standards (VRM, Unified, ARKit, SRanipal etc) internally, without futher configuration.

VRChat models have to be Unified or SRanipal, if I'm not wrong.

I believe iFacialMocap is similar to VMC, but it usually used to ARKit, as it's for iPhones.

At VSeeface, which I'm more familiar with, you can only adjust some thresholds and smoothings

[HelikeLong]

While I was finishing writing the reply, something just came to mind, there's an app called VRCF2VMC, I believe I can use it as a middle ground to mix FoxyFace to VMC and use in with my VTuber, will give it a try later (kinda really busy right now, probs will be able to test only next week)

[Plushie23131]

Any progress?

[Jeka8833]

I am working on improving face tracking, completely rewriting VRCFT and adding new and more accurate blendshape blending methods. This will also enable easy integration of, for example, VMC, without losing tracking quality. The current VRCFT implementation is highly dependent on the quality of the avatar, and the new implementation should provide a significant boost in tracking quality, especially on poorly made avatars.

I might be able to finish this, but it will take many months.

[Plushie23131]

oooh okok