LatexBlend: Scaling Multi-concept Customized Generation with Latent Textual Blending

2025-03-10

AI Summary

To address the degradation in generation quality and low computational efficiency in multi-concept customized text-to-image synthesis, this paper proposes the Implicit Text Space (ITS) framework. ITS constructs a learnable implicit representation space atop a frozen text encoder, enabling concept-specific modeling and compact storage in a shared “concept bank.” During inference, unbiased composition is achieved via linear projection and latent-feature-driven dynamic mixing. This work is the first to unify concept disentanglement, storage, and composition within an implicit space—supporting infinite concept scalability while preserving structural consistency and subject fidelity. Experiments demonstrate that ITS significantly outperforms baselines on multi-concept generation tasks, improving visual fidelity and compositional harmony. Moreover, it accelerates inference by 42% and reduces GPU memory consumption by 58%.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Natural Language Processing: Generation
• Machine Learning: Multimodal Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Customized text-to-image generation renders user-specified concepts into novel contexts based on textual prompts. Scaling the number of concepts in customized generation meets a broader demand for user creation, whereas existing methods face challenges with generation quality and computational efficiency. In this paper, we propose LaTexBlend, a novel framework for effectively and efficiently scaling multi-concept customized generation. The core idea of LaTexBlend is to represent single concepts and blend multiple concepts within a Latent Textual space, which is positioned after the text encoder and a linear projection. LaTexBlend customizes each concept individually, storing them in a concept bank with a compact representation of latent textual features that captures sufficient concept information to ensure high fidelity. At inference, concepts from the bank can be freely and seamlessly combined in the latent textual space, offering two key merits for multi-concept generation: 1) excellent scalability, and 2) significant reduction of denoising deviation, preserving coherent layouts. Extensive experiments demonstrate that LaTexBlend can flexibly integrate multiple customized concepts with harmonious structures and high subject fidelity, substantially outperforming baselines in both generation quality and computational efficiency. Our code will be publicly available.

Problem

• Scaling multi-concept customized text-to-image generation
• Improving generation quality and computational efficiency
• Enabling seamless blending of multiple concepts in latent space

Innovation

• Latent Textual space for concept blending
• Compact concept bank with latent features
• Reduced denoising deviation for coherent layouts