Every motion source, its licence, and which mode it is allowed in.
The motion she plays at home is cut from datasets released for research and non-commercial use, and from footage that may not be redistributed. None of it is displayed on this site, distributed, or reachable, and models trained on it never leave the machine she runs on. That is what "private forever" means: not a setting, a property of where the data can be. Her body from those models plays live and is never published or distributed, at home (private) and for friends on a call (public); the stream runs only on licence-clean motion, so nothing broadcast is derived from a source that forbids it.
Non-commercial data may play live, off the stream. It is never training input for anything public. A gate can refuse a clip; nothing can un-train a weight, so the line is drawn at the training set, not at the output.
All 35 sources in the registry, about 5220.6 hours listed, fewer unique: several overlap (HumanML3D is drawn from AMASS, BEAT v1 and BEAT2 are the same recordings, GENEA 2023 is built from Talking With Hands), with each licence as the source states it and the mode it is allowed in. "(by choice)" marks sources whose licence would allow more; they stay private because everything trained alongside them is.
| Source | Licence | Format | Size | Mode | Why it is here |
|---|---|---|---|---|---|
| Talking With Hands 16.2M (Meta) | CC BY-NC 4.0 | bvh+audio | 20 h | home and calls, never stream | two-person conversation with fingers |
| ZeroEGGS (Ubisoft) | CC BY-NC-ND 4.0 | bvh+audio | 2 h | home and calls, never stream | 19 styles, one actor; already partly in the corpus |
| BEAT2 / EMAGE (MPI) | CC BY-NC 4.0 | smplx+audio | 60 h | home and calls, never stream | the big one, 20 GB on HF; partly in the corpus already |
| BEAT v1 (BVH) | CC BY-NC 4.0 | bvh+audio | 76 h | home and calls, never stream | same recordings as BEAT2 in BVH with facial blendshapes |
| TalkSHOW / SHOW | MPI non-commercial research | smplx+audio | 27 h | home and calls, never stream | talk-show hosts fitted from video: real conversational gesture at a desk |
| Trinity Speech-Gesture | custom non-commercial research | bvh+audio | 4 h | home and calls, never stream | GENEA 2020/2022 base data |
| GENEA Challenge 2023 data (TWH-based, dyadic) | CC BY-NC 4.0 | bvh+audio+tsv | 18 h | home and calls, never stream | cleaned TWH with transcripts and the interlocutor |
| PATS (CMU, 2D poses of 25 speakers) | CC BY-NC 2.0 | 2d-keypoints+audio | 251 h | home and calls, never stream | huge but 2D; needs lifting before it is any use |
| TED Expressive (HA2G) | not stated (code GPL-3.0) | 3d-keypoints+audio | 27 h | home and calls, never stream | lifted TED talks; upper body only |
| Seamless Interaction (Meta) | CC BY-NC 4.0 | video+smplx codes | 4000 h | home and calls, never stream | 27 TB of tar shards; seamless_stream.py keeps everything but the video |
| DnD Group Gesture (Aalto) | not stated | bvh+audio | 6 h | home and calls, never stream | five-person tabletop conversation |
| Motion-X (IDEA) | IDEA non-commercial research | smplx | 144 h | home and calls, never stream | 81k sequences incl. in-the-wild video |
| AMASS (MPI) | MPI non-commercial research | smplx/smplh npz | 40 h | home and calls, never stream | the mocap library; HumanML3D is its labelled subset |
| HumanML3D | AMASS terms (code MIT) | amass subset + text | 29 h | home and calls, never stream | text labels for AMASS motions (needs AMASS) |
| LAFAN1 (Ubisoft) | CC BY-NC-ND 4.0 | bvh | 4.6 h | home and calls, never stream | locomotion, clean, five subjects |
| Bandai Namco Research Motiondataset | CC BY-NC-ND 4.0 | bvh | 3 h | home and calls, never stream | styled everyday motion incl. gestures |
| 100STYLE | CC BY 4.0 | bvh | 4 h | home and calls, never stream (by choice) | 100 walking styles; idle/character (CC BY 4.0) |
| CMU Motion Capture (BVH conversion) | free | bvh | 9 h | home and calls, never stream (by choice) | 2,500 clips, the classic (Hahne BVH port) |
| SFU Motion Capture | research only, no redistribution | bvh | 2 h | home and calls, never stream | clean everyday motion; per-clip .bvh links under nusmocap/ |
| KIT Motion-Language | not stated | bvh/mmm + text | 11 h | home and calls, never stream | text-described motions |
| Motorica Dance | custom non-commercial | bvh+audio | 6 h | home and calls, never stream | dance; rhythm and weight for the reflex layer |
| AIST++ | CC BY 4.0 (annotations) | smpl+audio | 5 h | home and calls, never stream | dance; SMPL motions 306 MB + 3D keypoints (terms of use accepted by pax for private use) |
| Learning2Listen (Ng et al. 2022) | not stated | DECA face+head coeffs + mel audio | 72 h | home and calls, never stream | LISTENER head and expression aligned to the speaker: her nods while pax talks |
| ViCo listening-head | research only, restricted | 3DMM coeffs + video | 2 h | home and calls, never stream | speaker and listener 3DMM in conversation |
| IEMOCAP (USC) | USC licence on request | face+head mocap markers + audio + emotion | 12 h | home and calls, never stream | dyadic face and head markers with emotion labels |
| VOCASET (MPI) | MPI non-commercial research | 4D face scans + audio | 1 h | home and calls, never stream | audio-driven mouth and jaw, the lipsync reference |
| MEAD emotional talking face | not stated | video, 60 actors, 8 emotions | 40 h | home and calls, never stream | expression under emotion at three intensities |
| HDTF talking heads | CC BY 4.0 (list; videos are YouTube) | YouTube video list + script | 16 h | home and calls, never stream | high-res talking heads; video via yt-dlp, not automated |
| CelebV-HQ | non-commercial research | YouTube video list + script | 68 h | home and calls, never stream | 35k face clips with expression/action labels; video via yt-dlp, not automated |
| Gaze360 (MIT) | non-commercial research | images + 3D gaze | n/a (poses) | home and calls, never stream | only if we ever train a gaze estimator; not animation data |
| InterHuman (InterGen) | CC BY-NC-SA 4.0 | smpl | 7 h | home and calls, never stream | two-person interaction |
| Inter-X | non-commercial research | smplx | 13 h | home and calls, never stream | two-person with hands |
| InterAct (2025) | not stated | smplx | 241 h | home and calls, never stream | daily two-person activities, large |
| Goliath-SC self-contact poses (ICCV 2025) | not stated (Goliath is CC BY-NC 4.0) | smplx poses | n/a (poses) | home and calls, never stream | 383K self-contact poses: exactly the fence's problem |
| TUCH / MTP self-contact (MPI) | MPI non-commercial research | smplx poses | n/a (poses) | home and calls, never stream | mimic-the-pose self-contact data |