README.md

January 23, 2026 · View on GitHub

Skywork-UniPic

Unified multimodal model for image editing, generation, and understanding

📝 Overview

Welcome to the Skywork-UniPic repository!
This repository hosts the model weights and official implementations of unipic unified multimodal series, featuring three distinct modeling paradigms:

  • UniPic-3 (README) — 🔥 Open-source SOTA Multi-Image Editing Model. Unified framework for single-image editing & multi-image composition. Supports 1–6 input images with flexible resolutions. 8-steps inference with 12.5× speedup via CM + DMD distillation.

    UniPic-3 Teaser

  • UniPic-2(README) — SD3.5M-Kontext and MetaQuery variants based on Efficient Architectures with Diffusion Post-Training, delivering state-of-the-art performance in text-to-image generation, fine-grained image editing, and multimodal reasoning.

    UniPic-2 Teaser

  • UniPic-1(README) — 1.5B parameters, Unified Autoregressive Modeling for joint visual understanding and generation, enabling a single transformer to handle both perception and synthesis tasks.

🔥 Latest News

DateUpdate
2026-01-09Released UniPic-3 — 🔥 Open-source SOTA multi-image editing model. Support single & multi-image editing, 1–6 inputs, 8-step / 12.5× faster inference
GitHub HuggingFace arXiv
2025-08-13Released UniPic-2 — Unified Model Weights with Diffusion-based Post-Training
GitHub HuggingFace arXiv
2025-07-30Released UniPic-1 — Autoregressive unified modeling from scratch
GitHub HuggingFace arXiv

✨ Key Features

  • 🎨 Text-to-Image Generation — High-fidelity synthesis from natural language prompts.
  • 🛠 Image Editing — Seamless inpainting, outpainting, and object manipulation.
  • 🖼 Image Understanding — Robust perception capabilities for various visual tasks.
  • ⚡ Efficient Architecture — Optimized for both accuracy and deployability.

📜 License

This project is licensed under the MIT License — see the LICENSE file for details.