Unified motion-language model for text-to-motion, motion captioning, completion and editing, from HKUST and Tencent Games.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Möbius is a research model for bidirectional human motion understanding and synthesis. A single architecture handles text-to-motion generation, motion-to-text captioning, motion completion (from a partial motion, with or without a text prompt) and motion editing. It was developed by researchers at HKUST, Tencent Games Motus and MIT.
The project page reports results on the HumanML3D test set. Two sizes are listed: a 0.6B model trained on HumanML3D only and a 1.7B model trained on a larger dataset. In text-to-motion, the 0.6B model reports R@1 of 0.8209 and FID of 6.13, compared with MotionStreamer (FID 12.21) and MotionGPT3 (FID 20.99). For motion captioning under the TM2T protocol, the 1.7B model reports the best scores among the compared methods, including CIDEr of 0.5836 on the raw captions.
At the time of the source material, the code is reported to be under internal review (per a YouTube creator summary), and no weights, repository or demo link is listed. The accompanying paper is cited as arXiv:2611.05279. It is aimed at motion researchers and animation or game developers, and currently exists as a project page and paper only.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Single architecture for text-to-motion generation, motion captioning, completion and editing, outperforming competitors on benchmarks. Code is under internal review.
Closest neighbours by meaning, not just by category.
Model
Text-to-human-motion flow-matching model that generates full-resolution motion directly, with training-free joint constraints.
Research
Diffusion framework that learns text-to-3D human motion generation from 2D poses extracted from monocular videos.
Research
Research method generating text-conditioned 3D human-object interaction motion using body-part anchors and a contact refiner.
Model · NVIDIA
NVIDIA autoregressive diffusion model for real-time, text- and constraint-controlled 3D human motion generation.
Model · Nvidia
Nvidia platform for generating character animations from text prompts, used as a source of reference motion.
Model
Text-prompted motion generation model that animates rigged 3D assets with arbitrary skeleton topologies, no retraining.