AesopAgent: Agent-driven Evolutionary System on Story-to-Video Production
Concetti Chiave
AesopAgent is an innovative agent-driven evolutionary system that converts user story proposals into high-quality videos through the integration of multiple generative capabilities within a unified framework.
Sintesi
Overview:
AesopAgent by DAMO Academy, Alibaba Group, focuses on multimodal content generation.
The system integrates generative capabilities for script, image, audio, and video assembly.
Horizontal Layer:
Utilizes RAG-based evolutionary system for video generation workflow optimization.
Includes script generation, image editing, and video assembly steps.
Utility Layer:
Provides utilities for consistent image generation in terms of composition, characters, and style.
Includes modules for image composition rationality, multiple characters consistency, and image style consistency.
Evaluation:
AesopAgent outperforms ComicAI and Artflow in visual storytelling ability and coherence.