LayerFusion: Harmonized Multi-Layer Text-to-Image Generation with Generative Priors

TL;DR

This paper introduces LayerFusion, a novel multi-layer image generation method using Latent Diffusion Models that produces more coherent layered images with dynamic interactions, improving quality and consistency.

Contribution

The paper presents a harmonized multi-layer generation approach for layered images using Latent Diffusion Models, enabling dynamic interactions between layers for better coherence.

Findings

Significant improvements in visual coherence and image quality.

Enhanced layer consistency over baseline methods.

Effective generation of transparent foreground and background layers.

Abstract

Large-scale diffusion models have achieved remarkable success in generating high-quality images from textual descriptions, gaining popularity across various applications. However, the generation of layered content, such as transparent images with foreground and background layers, remains an under-explored area. Layered content generation is crucial for creative workflows in fields like graphic design, animation, and digital art, where layer-based approaches are fundamental for flexible editing and composition. In this paper, we propose a novel image generation pipeline based on Latent Diffusion Models (LDMs) that generates images with two layers: a foreground layer (RGBA) with transparency information and a background layer (RGB). Unlike existing methods that generate these layers sequentially, our approach introduces a harmonized generation mechanism that enables dynamic interactions between the layers for more coherent outputs. We demonstrate the effectiveness of our method through extensive qualitative and quantitative experiments, showing significant improvements in visual coherence, image quality, and layer consistency compared to baseline methods.