Views
No views yet
nvidia/ARDY-Core-RP-20FPS-Horizon8,
released by TREE Industries for responsive real-time motion generation.| Property | Value |
|---|---|
| Skeleton | ARDY Core, 27 joints |
| Motion rate | 20 FPS |
| Generation horizon | 8 frames |
| Frames per token | 4 |
| Maximum runtime window | 64 tokens / 256 frames |
| Latent dimension | 128 |
| Text feature dimension | 4096 |
| Diffusion steps | 10 |
| Source revision | 257a0843a10bf5201065963d7bca2791e9393a7a |
encoder.onnx: normalized motion history to FSQ latent tokensdenoiser.onnx: guided, text-conditioned latent motion denoisingdecoder.onnx: latent tokens to local-root and body motion featuresmodel_contract.json records the complete skeleton and tensor contract.fp32/ contains exact FP32 reference exports.directml-mixed-fp16/ contains compact runtime builds with FP32 graph I/O
and sensitive operations plus selected FP16 constant-weight matrix islands.TREEIndustries/Llama-3-ARDY-Text-Encoder-ONNX.onnx.checker. The mixed-FP16 graphs were executed
against FP32 with identical deterministic 64-token/256-frame inputs using ONNX
Runtime 1.20.1 and DirectML. Shapes matched and every output was finite.
Measured differences are in VALIDATION.md.Licensed by NVIDIA Corporation under the NVIDIA Open Model License
1@article{zhao2026ardy,
2 title = {ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation},
3 author = {Zhao, Kaifeng and Petrovich, Mathis and Zhang, Haotian and Wang, Tingwu and Tang, Siyu and Rempe, Davis},
4 journal = {ACM Transactions on Graphics (TOG)},
5 year = {2026},
6 volume = {45},
7 number = {4},
8 articleno = {86},
9 doi = {10.1145/3811284}
10}