itch.io

I cloned the official OpenSeeFace source and ran its native Python tracker under Linux. Using Python 3.10 with the required dependencies, this command works:

python facetracker.py -c 0 -W 1280 -H 720 -F 30

It successfully accesses /dev/video0, detects and tracks the face, and sends tracking data to the default OpenSeeFace destination: 127.0.0.1:11573

With this running, vTuber Kit immediately responds to my face. The webcam and Unity OpenSee UDP receiver both work correctly on Linux. The problem appears isolated to the OpenSeeLauncher/bundled OpenSeeFace tracker portion of the Linux build. 

The existing Unity OpenSee receiver already works on Linux, so the simplest fix may be to provide a Linux-compatible OpenSeeFace tracker and have the Linux build launch that instead of the Windows facetracker.exe. No changes to the vTuber Kit UDP receiver were required for this workaround.

Hi! I’m testing the Linux build of vTuber Kit v0.2.2 on CachyOS/KDE and found what appears to be an OpenSeeFace/camera initialization issue.

**System:**

* OS: CachyOS (Arch Linux), KDE

* Laptop: HP ENVY x360 Convertible 15m-ed0xxx

* Camera: HP Wide Vision HD Camera

* Integrated GPU: Intel Iris Plus Graphics (ICL GT2)

* Mesa: 26.2.1

* Unity: 2021.3.45f2

The Linux build launches successfully after fixing one filename case-sensitivity issue:

`Custommouth2.png` was packaged, while the application expected `CustomMouth2.png`.

The webcam itself works correctly under Linux. `v4l2-ctl` reports:

`/dev/video0` — MJPEG up to 1280x720 @ 30 FPS and YUYV up to 640x480 @ 30 FPS.

I also successfully tested streaming directly from the camera with:

`v4l2-ctl --stream-mmap --stream-count=100 -d /dev/video0`

and it captured at approximately 30 FPS.

However, selecting any of the camera options (Option A, B, or C) in vTuber Kit does not open `/dev/video0` or `/dev/video1`. Checking with `lsof /dev/video0 /dev/video1` returns nothing.

The Unity `Player.log` shows this error when attempting to list/start the camera:

`EntryPointNotFoundException: CreateJobObject`

with the relevant stack:

`OpenSee.Job.CreateJobObject`

`OpenSee.OpenSeeLauncher.CheckSetup`

`OpenSee.OpenSeeLauncher.ListCameras`

and when starting tracking:

`OpenSee.OpenSeeLauncher.StartTracker`

`OpenSee.OpenSeeLauncher.CameraToggle`

I also inspected the bundled OpenSeeFace directory. It contains Windows components such as:

* `facetracker.exe`

* `dshowcapture_x64.dll`

* `dshowcapture_x86.dll`

* `escapi_x64.dll`

* `escapi_x86.dll`

* `python37.dll`

* multiple `win_amd64.pyd` files

but there are no OpenSeeFace Linux `.so` or Linux executable files present. The only `.bin` found under OpenSeeFace is the model `models/benchmark.bin`.

This makes me suspect the Linux build is currently attempting to use the Windows OpenSeeFace backend, or that the Linux-specific OpenSeeFace component is missing/not being loaded correctly.

The webcam itself appears to be fully functional under Linux, so I don't think this is a V4L2/driver issue.

Hopefully this helps track down the Linux camera/tracking issue. 

Hey Y'all & Kyuppin! I could've sworn you could map Custom Objects to the mouth like how you can so the eyes, nose, face, and other various parts of the model. I was wondering if this was never a thing with the mouth or if this is just something I'm not able to do. If you got any answers for me that'd be great thanks!

Anyway in the future/upcoming updates that the software will be able to allow the user to adjust the duration of wildcard via slider like some other setting in the app?

Do you think a redemption like what TITS has for multiple objects will be possible through the program? I've been wanting my Bonk to have a multi-redeemable option cuz I think it'd be funny. 

Is there a possibility of adding a TTS feature for a sort of pet or custom Item? Was thinking It'd be cool to have an item grow like a face or something when triggered by TTS.