The authors designed CineEvolve as a unified vision-language policy that performs two roles: a Director for initial creation and targeted revision, and a Critic for defect diagnosis.
Self-evolving AI agent generates cinematic multi-shot videos by iteratively directing and critiquing itself
The CineEvolve framework uses a unified vision-language model that learns to create, diagnose, and revise video sequences through alternating reinforcement learning.
Chinese Tech
Jingzhe Zhang · Peixuan Zhang · Genying Li · Bing Ma · Shuchen Weng · Si Li · +1 more
Beijing University of Posts and Telecommunications · Peking University · Beijing Academy of Artificial Intelligence · Tencent
Research Digest··2 min read
The authors propose CineEvolve, a framework where a single vision-language policy acts as both Director and Critic to generate cinematic multi-shot videos.
§