What do ARKit and MediaPipe Face Landmarker document about blendshape count, value range, left/right meaning, coordinate frames and frame timing?
Partially verified: the curator re-read S27, S29, S30; S28, S31, S32 were not re-read by the curator and rest on the producer's pinned reading. ARKit and MediaPipe both expose 52 blendshape coefficients; ARKit groups its constants (eyes 7+7, mouth/jaw 27, brows/cheeks/nose 10, tongue 1) in a right-handed metre face space, and MediaPipe's Left/Right names refer to the subject, so a mirrored (selfie) input swaps them. Neither document states that the names or semantics map one-to-one.
Partially verified: the curator re-read S27, S29, S30; S28, S31, S32 were not re-read by the curator and rest on the producer's pinned reading.
ARKit and MediaPipe both expose 52 blendshape coefficients; ARKit groups its constants (eyes 7+7, mouth/jaw 27, brows/cheeks/nose 10, tongue 1) in a right-handed metre face space, and MediaPipe's Left/Right names refer to the subject, so a mirrored (selfie) input swaps them. Neither document states that the names or semantics map one-to-one.
Curator qualifications: Partially verified: the curator re-read S27, S29, S30; S28, S31, S32 were not re-read by the curator and rest on the producer's pinned reading. Curator GET 2026-10-23 of the sources not re-read (mechanical source evidence, not semantic verification): S28: HTTP 200, bytes equal to the producer pin; S31: HTTP 200, bytes differ from the producer pin; S32: HTTP 200, bytes equal to the producer pin. Docs read 2026-10-22; no face capture or retargeting run. Blendshape V2 model card PDF over the readback cap (item-level hold); Epic Live Link Face page unreadable (JavaScript) per producer; MediaPipe VIDEO-mode timestamp claim from producer reading.
Limitations: Docs read 2026-10-22; no face capture or retargeting run.
Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.
Structured fields
{
"citations": [
{
"url": "https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor.json",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
},
{
"url": "https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor/blendshapes.json",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
},
{
"url": "https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor/blendshapelocation.json",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
},
{
"url": "https://ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
},
{
"url": "https://ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker/python",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
},
{
"url": "https://storage.googleapis.com/mediapipe-assets/Model%20Card%20Blendshape%20V2.pdf",
"note": "Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live)"
}
]
}Evidence claims
- [sourced_fact] ARKit and MediaPipe both expose 52 blendshape coefficients; ARKit groups its constants (eyes 7+7, mouth/jaw 27, brows/cheeks/nose 10, tongue 1) in a right-handed metre face space, and MediaPipe's Left/Right names refer to the subject, so a mirrored (selfie) input swaps them. Neither document states that the names or semantics map one-to-one. · locator: Apple docs JSON arfaceanchor/blendshapelocation (topicSections); Apple docs JSON arfaceanchor (transform discussion); MediaPipe Face Landmarker task page (live) source
Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| documentation | unknown | permitted | permitted | unknown | unknown | Applies to this original summary only, not to the cited sources. |
Linked records
- cites → https://ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker
- cites → https://ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker/python
- cites → https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor.json
- cites → https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor/blendshapelocation.json
- cites → https://developer.apple.com/tutorials/data/documentation/arkit/arfaceanchor/blendshapes.json
- cites → https://storage.googleapis.com/mediapipe-assets/Model%20Card%20Blendshape%20V2.pdf