face stack: YuNet (MIT) + MediaPipe Face Landmarker v2 + blendshapes (Apache 2.0) replace the non-commercial InsightFace pair

README.md

@@ -16,7 +16,8 @@ Curated model bundle for [Pop Maker Studio](https://github.com/verticalrectangle
 | Moondream2 vision | [Xenova/moondream2](https://huggingface.co/Xenova/moondream2) | Apache 2.0 |
 | Piper voices | [rhasspy/piper-voices](https://huggingface.co/rhasspy/piper-voices) | MIT |
 | RobustVideoMatting (mobilenetv3) | [PeterL1n/RobustVideoMatting](https://github.com/PeterL1n/RobustVideoMatting) | GPL-3.0 |
-| SCRFD-10G + 2d106det (face) | [InsightFace model zoo](https://github.com/deepinsight/insightface) | non-commercial research (see upstream) |
+| YuNet (face detection) | [OpenCV Zoo](https://github.com/opencv/opencv_zoo/tree/main/models/face_detection_yunet) | MIT |
+| Face Landmarker v2 + blendshapes | [MediaPipe](https://ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker) (tflite→onnx via tf2onnx) | Apache 2.0 |
 | RVC pitch bases | generated from [lj1995 pretrained_v2](https://huggingface.co/lj1995/VoiceConversionWebUI) f0G emb_pitch tensors | MIT (derivative) |
 
 Layout note: the app expects these under `build/models/` — flat names for the

face/{2d106det.onnx → face_blendshapes.onnx}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f001b856447c413801ef5c42091ed0cd516fcd21f2d6b79635b1e733a7109dbf
-size 5030888
+oid sha256:6d7f5b91e5ed881040883d21c908ef5a376d82891de7c22c1d9cc34f330108b3
+size 1883546

face/{scrfd_10g.onnx → face_landmarks_v2.onnx}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5838f7fe053675b1c7a08b633df49e7af5495cee0493c7dcf6697200b85b5b91
-size 16923827
+oid sha256:deb88b9cc49a59e6a9ceacd91e1d91fe0c3cf44ba1fb77ecf1e5633ef54f9271
+size 4920995

face/yunet.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f2383e4dd3cfbb4553ea8718107fc0423210dc964f9f4280604804ed2552fa4
+size 232589