UniMate, a research project from Princeton, UC Berkeley, MIT and NTU, has released training and inference code under the MIT license plus preview checkpoints under CC BY-NC 4.0 for a unified text-to-animation model accepted to SIGGRAPH Asia 2026. It also released UniML3D, described as 13,769 text-paired motion clips across 7,430 skeletons spanning bipedal, quadrupedal, avian, marine, insectoid and serpentine rigs as well as articulated rigid objects. The README says inference generates motion for arbitrary rigged assets in real time without per-skeleton retraining or test-time optimization, and an October 5 rig_preprocess tool converts GLB, glTF or FBX assets; the authors acknowledge that many motions and skeletons still fail.