Generative Motion Infilling from Imprecisely Timed Keyframes Volume: 44 Issue: 2
P. Goel 1 H. Zhang 2 C. K. Liu 1 K. Fatahalian 1 1 Stanford University 2 NVIDIA · 2025
Read Generative Motion Infilling from Imprecisely Timed Keyframes Volume: 44 Issue: 2 by P. Goel 1 H. Zhang 2 C. K. Liu 1 K. Fatahalian 1 1 Stanford University 2 NVIDIA (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01016.txt.
Description
Purvi Goel, Haotian Zhang, Karen Liu, and Kayvon Fatahalian animate movement from keyframes whose timing is only roughly right, in a 2025 paper. Animators pose keyframes and let software fill the gaps, but learned fillers demand exact timing. Hand-setting perfect timings is tedious, and small errors degrade the motion.
Their model fixes timing as it generates. The architecture outputs a time-warping function that retimes the constraints, plus spatial residuals that add pose detail. Trained this way, the system takes approximately timed keyframes and returns varied, realistic motion with believable pacing and fine submovements. The authors show the retiming happening automatically inside the generation process. The work targets animators and game developers who think in poses, not stopwatches. Sloppy timing, long a source of uncanny results, becomes an acceptable input.
Catalog record
| CREATOR | P. Goel 1 H. Zhang 2 C. K. Liu 1 K. Fatahalian 1 1 Stanford University 2 NVIDIA |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01016