Exploring MLLM-Diffusion Information Transfer with MetaCanvas
This addresses the gap between multimodal understanding and generation for applications requiring precise visual control, representing a novel method rather than an incremental improvement.
The paper tackled the problem of multimodal large language models (MLLMs) being underutilized as global text encoders in visual generation, limiting precise control, and proposed MetaCanvas, a lightweight framework that enables MLLMs to reason in latent spaces, resulting in consistent outperformance over baselines across six tasks like text-to-image and video generation.
Multimodal learning has rapidly advanced visual understanding, largely via multimodal large language models (MLLMs) that use powerful LLMs as cognitive cores. In visual generation, however, these powerful core models are typically reduced to global text encoders for diffusion models, leaving most of their reasoning and planning ability unused. This creates a gap: current multimodal LLMs can parse complex layouts, attributes, and knowledge-intensive scenes, yet struggle to generate images or videos with equally precise and structured control. We propose MetaCanvas, a lightweight framework that lets MLLMs reason and plan directly in spatial and spatiotemporal latent spaces and interface tightly with diffusion generators. We empirically implement MetaCanvas on three different diffusion backbones and evaluate it across six tasks, including text-to-image generation, text/image-to-video generation, image/video editing, and in-context video generation, each requiring precise layouts, robust attribute binding, and reasoning-intensive control. MetaCanvas consistently outperforms global-conditioning baselines, suggesting that treating MLLMs as latent-space planners is a promising direction for narrowing the gap between multimodal understanding and generation.