# emilianavt/OpenSeeFace

Robust realtime face and facial landmark tracking on CPU with Unity integration

Repository: https://github.com/emilianavt/OpenSeeFace
Canonical: https://ross.abutalabs.com/products/openseeface
Language: Python
License: BSD-2-Clause
License Family: permissive
Topics: face-tracking, face-landmarks, depth-estimation, unity, unity3d, python, csharp, udp, onnx, onnxruntime, virtual-youtuber, vtuber, mobilenetv3, pytorch, openseeface, face-detection, detection-model, landmark-model, tracker, cpu
Last push: 2025-12-28T11:52:30+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 59, release rhythm 8, longevity 100
- inputs: {"age_days": 2436, "days_push": 248, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2038, forks 205 (observed 2026-08-28T04:06:08.242874+00:00)

## What it is
OpenSeeFace is a robust realtime face and facial landmark tracking library that runs on CPU at 30-60 fps using ONNX-converted MobileNetV3 models. It streams tracking data over UDP and includes Unity integration for animating avatars, powering tools like VSeeFace and VTube Studio.

## Use cases
- track facial landmarks from a webcam in realtime on CPU
- animate a VRM avatar with webcam face tracking in Unity
- build a vtuber puppeteering app with head pose and expression tracking
- get stable face landmark detection in low light or noisy video
- stream face tracking data over UDP to another application
- compare face tracking robustness against MediaPipe

## When to choose
- you need realtime face tracking on CPU without a GPU
- you want stable landmarks under poor lighting, noise, or extreme head poses
- you are building a vtuber or avatar animation tool with Unity or Godot integration
- you need a lightweight tracking library rather than a full end-user application

## When to avoid
- you want a ready-made avatar puppeteering program - use VSeeFace or VTube Studio instead
- you need highly accurate eye-region tracking
- you require landmarks matching standard iBUG 68 conventions exactly
- you need GPU-accelerated multi-face tracking at scale

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, sdk
- domain: computer-vision, machine-learning, graphics, developer-tools
- platform: python, windows, cross-platform
- tags: face-tracking, facial-landmarks, vtuber, onnx, unity, udp, realtime, cpu-inference, gpu

## Member repositories
- emilianavt/OpenSeeFace (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:08.242874+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:59:14.562976+00:00, confidence not recorded.
  - readme: https://github.com/emilianavt/OpenSeeFace (fetched 2026-08-28T04:06:08.242874+00:00, sha a6d43d96ef3f)
- Data as of 2026-08-30T08:39:29.467469+00:00.