ComfyUI nodes that crop the masked area before sampling and stitch the inpainted result back, making inpainting faster and more controllable.
scraed/LanPaint
A training-free inpainting and outpainting sampler for ComfyUI that works with many diffusion and rectified-flow image and video models.
Open-source alternative to:Photoshop Generative Fill
- Stars
- 1,466
- Forks
- 55
- Open issues
- 22
- Last commit
- 24 days ago
What is it?
LanPaint is a sampler for ComfyUI that performs mask-constrained inpainting and outpainting on top of existing pretrained diffusion and rectified-flow models. It requires no fine-tuning or backpropagation. Its distinguishing feature is a "thinking" mode: the sampler runs several iterations on each denoising step, and the user can trade extra computation time for better inpainting quality.
The project's nodes are designed to slot into a standard ComfyUI workflow in place of the default KSampler. The repository lists support for a wide range of image models, including Flux, SD 3.5, SDXL, Qwen Image and Edit, Wan 2.2, Z-Image, and others. It also covers video inpainting and outpainting with Wan 2.2, plus a video mask editor and an audio-video pipeline for MiniMax H3.
The repository includes a paper published at TMLR 2025, a Diffusers implementation maintained separately, and example workflows for each supported model. Note that the project warns about degraded results on distilled models such as Flux.dev and recommends keeping mask values binary.
Key features
- 01Training-free inpainting and outpainting that works with existing diffusion and rectified-flow models.
- 02"Thinking" iterations per denoising step let users trade compute time for quality.
- 03Drop-in replacement for the default ComfyUI KSampler, with basic and advanced variants.
- 04Video inpainting and outpainting with Wan 2.2, plus a per-frame video mask editor.
- 05Joint video and audio inpainting for the MiniMax H3 model.
Use cases
- Removing or replacing objects and regions in generated or photographed images.
- Extending the canvas of an image or video with outpainting.
- Making targeted edits to a video across frames while keeping motion consistent.
- Generating consistent character views by inpainting regions of an existing image.
Quick start
- 1
Install ComfyUI following the official guide
- 2
Install ComfyUI-Manager for extension management
- 3
Install LanPaint nodes via ComfyUI-Manager by searching for LanPaint
- 4
Restart ComfyUI to load the LanPaint nodes
Things to know
- Runs only inside ComfyUI, so it is not a standalone tool.
- Performance is degraded on distillation models such as Flux.dev, and low Flux guidance is recommended.
- Licensed under GPL-3.0, a copyleft license that requires derivative works to be open source.
Topics
FAQ
- What is LanPaint?
- A training-free inpainting and outpainting sampler for ComfyUI that works with many diffusion and rectified-flow image and video models. LanPaint is a sampler for ComfyUI that performs mask-constrained inpainting and outpainting on top of existing pretrained diffusion and rectified-flow models. It requires no fine-tuning or backpropagation. Its distinguishing feature is a "thinking" mode: the sampler runs several iterations on each denoising step, and the user can trade extra computation time for better inpainting quality.
- Is LanPaint free to use? What license does it have?
- LanPaint is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
- Is LanPaint still maintained?
- Actively maintained. The last commit to LanPaint was on Sep 28, 2026.
- What can I use LanPaint for?
- Removing or replacing objects and regions in generated or photographed images. Extending the canvas of an image or video with outpainting. Making targeted edits to a video across frames while keeping motion consistent. Generating consistent character views by inpainting regions of an existing image.
- How do I install LanPaint?
- Install ComfyUI following the official guide Install ComfyUI-Manager for extension management Install LanPaint nodes via ComfyUI-Manager by searching for LanPaint Restart ComfyUI to load the LanPaint nodes
- Where can I get LanPaint?
- The source code and installation instructions are on GitHub: https://github.com/scraed/LanPaint
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A Krita plugin that brings generative AI image creation, inpainting and outpainting into your painting workflow, backed by a local or cloud ComfyUI server.
A web-based creative engine for generating and editing images with Stable Diffusion and other AI models, featuring a unified canvas and node-based workflows.
Open-source studio and portable .char file format for keeping AI characters consistent across image and video models, locally or via hosted APIs.
A Gradio-based web UI for running Stable Diffusion and newer image/video models, continuing the Forge fork with added model support and optimizations.
Offline, open-source image generator built on Stable Diffusion XL that lets you focus on prompts instead of manual parameter tuning.