PrismLayers: Open Data for High-Quality Multi-Layer Transparent Image Generative Models

ICLR 2026 Conference Submission14552 Authors

Keywords: Graphic Design; Multi-Layer Transparent Image; Diffusion Model

Abstract: Generating high-quality, multi-layer transparent images from text prompts can unlock a new level of creative control, allowing users to edit each layer as effortlessly as editing text outputs from LLMs. However, the development of multi-layer generative models lags behind that of conventional text-to-image models due to the absence of a large, high-quality corpus of multi-layer transparent data. We address this fundamental challenge by: (i) releasing four open, ultra-high-fidelity datasets—\multilayertrain, \multilayertrainplus, \multilayertrainpro, and \multilayerreal—consisting of 200K, 100K, 20K, and 1K multi-layer transparent images with accurate alpha mattes, respectively. (ii) introducing a training-free synthesis pipeline that generates such data on demand using off-the-shelf diffusion models, and (iii) delivering a strong multi-layer generation model, ART+, which matches the aesthetics of modern text-to-image generation models. The key technical contributions include: LayerFLUX, which excels at generating high-quality single transparent layers with accurate alpha mattes, and MultiLayerFLUX, which composes multiple LayerFLUX outputs into complete images, guided by human-annotated semantic layout. To ensure higher quality, we apply a rigorous filtering stage to remove artifacts and semantic mismatches, followed by human selection. Fine-tuning the state-of-the-art ART model on our synthetic \multilayertrainpro yields ART+, which outperforms the original ART in 60\% of head-to-head user study comparisons and even matches the visual quality of images generated by the FLUX.1-[dev] model. Our work establishes a solid dataset foundation for multi-layer transparent image generation, enabling research and applications that require precise, editable, and visually compelling layered imagery.

Primary Area: datasets and benchmarks

Submission Number: 14552