MMCORE: MultiModal COnnection with Representation Aligned Latent Embeddings
This work addresses the problem of efficient and high-quality multimodal image synthesis and editing for AI applications, representing an incremental improvement by streamlining existing methods.
The paper tackles multimodal image generation and editing by introducing MMCORE, a framework that uses a Vision-Language Model to predict semantic embeddings for conditioning a diffusion model, resulting in high-fidelity synthesis with reduced computational overhead and outperforming state-of-the-art baselines across various benchmarks.
We present MMCORE, a unified framework designed for multimodal image generation and editing. MMCORE leverages a pre-trained Vision-Language Model (VLM) to predict semantic visual embeddings via learnable query tokens, which subsequently serve as conditioning signals for a diffusion model. This streamlined design effectively transfers the rich understanding and reasoning capabilities of VLMs into the visual generation process. By obviating the need for deep fusion between autoregressive and diffusion models or training from scratch, MMCORE significantly reduces computational overhead while maintaining high-fidelity synthesis. MMCORE seamlessly integrates text-to-image synthesis with interleaved image generation, demonstrating robust multimodal comprehension in complex scenarios such as spatial reasoning and visual grounding. Comprehensive evaluations indicate that MMCORE consistently outperforms state-of-the-art baselines across a broad spectrum of text-to-image and single/multi-image editing benchmarks.