ChromaGS enables real-time language color editing of 3D avatars
ChromaGS: Text-Driven Semantic Editing of 4D Gaussian Avatars
Computer Vision and Pattern Recognition
Summary
Changing the colors of 3D animated heads usually takes a lot of work or retraining. The authors made a system called ChromaGS that lets users instantly change the colors of specific parts of 3D head avatars by simply typing instructions in natural language. Their method breaks down colors into main region colors and small details, so color changes stay neat and accurate without losing detail. It works quickly without needing new training, giving users exact control without messing up other parts.
What this means in practice
- •For 3d character artists: Modify colors of animated 3D head avatars instantly by typing color changes in natural language without retraining models.
- •For interactive media developers: Give users real-time control over facial color customization in avatars using simple text commands during rendering.
- •For game developers: Implement precise text-driven color editing of character head regions during gameplay without affecting other appearance details.$Commercial implications: Enables new avatar customization features sold as in-game enhancements and interactive character design tools.
Authors
Antonio Canela, Jordi Sànchez-Riera
Abstract
We present ChromaGS, a method for real-time, language-guided color editing of animatable 3D Gaussian head avatars. Given a trained animatable avatar, users can instantly modify the color of semantic regions through natural language, with edits applied at render time and no retraining required. Our key insight is to augment each Gaussian primitive with learned soft assignments to semantic regions and decompose colors into region-level base colors and Gaussian-level residuals. This decomposition enables coherent color transfer: modifying a region's base color propagates naturally through all associated Gaussians while preserving fine appearance details encoded in residuals. A two-stage language pipeline translates text instructions into target colors, supporting both absolute specifications and relative adjustments. Unlike generative editing methods that may introduce unintended modifications, our approach provides deterministic, precisely localized semantic control. Experiments demonstrate faithful appearance preservation and intuitive interaction across diverse subjects. Project page and code are available at: https://a-canela.github.io/chromags/