Boogu-Image-0.1-Edit: Open-Source Unified Image Editing with Apache 2.0

Boogu-Image-0.1-Edit is the image-to-image editing variant of the Boogu-Image-0.1 model family, released on June 16, 2026 under the permissive Apache 2.0 license. It provides instruction-based image editing capabilities powered by a unified multimodal understanding and generation (MLLM) backbone.

Overview

Boogu-Image-0.1 is a competitive open-source unified image generation and editing model family developed by the Boogu project. The family includes three main variants: Base (text-to-image), Turbo (4-step distilled fast generation), and Edit (image-to-image editing): all released under the Apache 2.0 license.

The Edit variant specifically focuses on instruction-based image editing: users provide a reference image along with a natural language instruction describing the desired edit, and the model generates the edited result while preserving the original image's structure and content.

Boogu-Image-0.1 achieved competitive results in the Boogu Arena (an LM Arena-style preference evaluation), ranking favorably against both closed-source systems and leading open-source alternatives across 1K+ diverse test prompts.

Key Features

Feature	Description
Task	Instruction-based image-to-image editing
Architecture	Unified MLLM understanding + diffusion generation
License	Apache 2.0 (fully open-source)
Library	Diffusers (custom `BooguImagePipeline`)
Languages	English and Chinese optimized
ComfyUI	Native support in ComfyUI

Model Architecture

Boogu-Image-0.1 employs a unified multimodal understanding and generation architecture that integrates:

A multimodal large language model (MLLM) for understanding user instructions and image content
A diffusion transformer for high-quality image generation
A VAE for latent space encoding/decoding

This unified approach allows the model to achieve precise instruction following while maintaining high image quality. The Edit variant specifically leverages the MLLM's understanding of spatial relationships, object attributes, and editing instructions to produce coherent modifications.

Capabilities

Boogu-Image-0.1-Edit excels at a variety of image editing tasks:

Object replacement: Swap objects in an image based on text descriptions
Background changes: Modify backgrounds while preserving foreground subjects
Style transfer: Apply artistic styles to existing images
Local edits: Modify specific regions guided by text instructions
Bilingual support: Handles both English and Chinese editing instructions

ComfyUI Integration

Boogu-Image-0.1-Edit is natively supported in ComfyUI. Get started quickly with the official Boogu Image Edit workflow.

Make sure you have updated ComfyUI to the latest version (update guide). The required model weights are available in the Comfy-Org/Boogu-Image repository on Hugging Face.

The Edit variant currently supports single-reference-image editing at 1K resolution. Multi-reference image support is planned for future releases.

Online Demos

You can try Boogu-Image-0.1-Edit directly in your browser:

Edit Demo: demo-edit.boogu.org
Base Demo: demo-base.boogu.org
Turbo Demo: demo-turbo.boogu.org

Availability

Hugging Face (Edit): Boogu/Boogu-Image-0.1-Edit
Hugging Face (Base): Boogu/Boogu-Image-0.1-Base
GitHub: boogu-project/Boogu-Image
Project Page: boogu.org
Gallery: boogu-gallery.netlify.app

The Boogu team does not currently provide any paid API or commercial service. Any paid product offered under the name "Boogu-Image" is not affiliated with this project.

Summary

Boogu-Image-0.1-Edit brings competitive instruction-based image editing to the open-source community under a permissive Apache 2.0 license. With its unified MLLM architecture, strong bilingual support, and out-of-the-box ComfyUI integration, it represents a significant step forward for open-source image editing tools.