Cygames and Tokyo Institute of Science Land ECCV2026 Paper on Motion Style Slider
The acceptance puts Cygames' research arm inside a top-tier computer vision venue, and the technique targets a specific production cost: adjusting an acting performance's intensity without extra motion capture.
Reporting from 1 source: GameBusiness.jp.
Cygames and Tokyo Institute of Science had a joint research paper accepted at ECCV2026, the computer vision conference held September 8-12 in Malmo, Sweden. The paper, "Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion," proposes a technique that adjusts acting intensity continuously like a slider, using data from Cygames' Osaka motion capture studio. The acceptance rate was 27.5 percent. The paper is registered as a poster presentation for September 11.
The method takes two motion clips as input: a neutral performance with no emotional expression and one with an acting style applied. From those two endpoints it can dial intensity up or down continuously, with no intermediate takes prepared in advance.
That answers a production problem the authors describe. Conventional style conversion handles expressions like joy, sadness, and anger, but fine-tuning their strength has meant extra motion capture or added data preparation, which costs time and money. The slider removes that step.
The work draws on motion data recorded at Cygames' Osaka motion capture studio. The eight authors span Tokyo Institute of Science, Cygames, and the University of Tokyo. Cygames is looking toward games, video, animation, and virtual character production, and toward integrating the method into production workflows.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.
Sources
- GameBusiness.jp サイゲームスと東京科学大学の共同研究がECCV2026に採択―3Dキャラの演技スタイルを無段階調整する新技術