Yume-1.5: Text-Controlled Interactive World Generation
Published:Dec 26, 2025 17:52
•1 min read
•ArXiv
Analysis
This paper addresses limitations in existing diffusion model-based interactive world generation, specifically focusing on large parameter sizes, slow inference, and lack of text control. The proposed framework, Yume-1.5, aims to improve real-time performance and enable text-based control over world generation. The core contributions lie in a long-video generation framework, a real-time streaming acceleration strategy, and a text-controlled event generation method. The availability of the codebase is a positive aspect.
Key Takeaways
Reference
“The framework comprises three core components: (1) a long-video generation framework integrating unified context compression with linear attention; (2) a real-time streaming acceleration strategy powered by bidirectional attention distillation and an enhanced text embedding scheme; (3) a text-controlled method for generating world events.”