AI-assisted inpainting and generation
Select an area or draw boxes on an image or empty canvas, describe what should be added, retained, moved or removed, and receive the result as a new layer.
Scumble
Scumble is a free, GPL-3.0 open-source desktop image editor for AI-assisted inpainting, editing, upscaling and image generation. It helps artists and image editors create model-generated changes as colour-matched, editable layers using ComfyUI, Comfy Cloud or supported provider APIs.
AI-assisted editing with editable layers
Scumble is a free, open-source desktop image editor built around AI models and ComfyUI. It is designed for inpainting, editing, upscaling and generation: users select part of an image or draw boxes on an empty canvas, describe the intended result, and receive the output as a separate layer. The layer can be masked, blended, filtered and adjusted before the document is flattened.
The application can render through a local or remote ComfyUI, Comfy Cloud, or supported provider APIs using the user's own keys. It also remains useful without an AI rendering connection for conventional editing, selection, painting, filters, saving and export. The project is currently in the 0.1.x stage, so the documentation advises users to expect rough edges and keep backups of their images.
Key features
Non-destructive layered editing
Generated layers are colour-matched around their edges and can be adjusted later with a Match slider. Layers can also be masked, erased, blended and filtered before the document is flattened.
Multiple rendering backends
Run recipes through a local or remote ComfyUI, use Comfy Cloud, or connect directly to provider APIs with keys for services including Black Forest Labs, OpenAI, Google, fal.ai, Replicate and OpenRouter.
Recipe-based ComfyUI workflows
On ComfyUI, workflows containing the Inpaint Canvas node can be used as recipes. The editor can open ComfyUI beside the workspace so users can modify and save a recipe graph.
Image processing tools
The editor includes GPU filters such as film looks, grain, curves and LUTs, along with in-app object selection, background removal and Remove tools that can run locally after their helper models are downloaded.
Automation and production files
Commands are exposed through an MCP server, allowing compatible clients to operate the editor. PSD and OpenRaster preserve layers, and a tile engine supports work on large documents without freezing the window as readily as a fully in-memory workflow.
Practical use cases
Photo retouching and object replacement
Open a photo, mark an unwanted object or an area that needs changing, describe the replacement, and keep the generated patch separate so it can be masked or blended before export.
Image creation from a layout
Start with a blank canvas, draw boxes to define where elements belong, and use a supported image model to create a composition whose generated parts arrive as editable layers.
Reference-based compositing
Place reference images as named layers and use model-specific references such as @img1 and @img2 when combining visual material or directing an edit.
Custom ComfyUI pipelines
Use a local workstation or rented remote machine running ComfyUI, select a recipe, and edit the recipe's graph when a controlled custom workflow is more suitable than a direct API call.
Assistant-driven image operations
Connect an MCP client to Scumble to open documents, make selections, generate edits, apply matching and export while retaining the editor's window and undo context.
Pros and Cons
Pros
- Generated results remain separate, editable layers with masks, blend modes, filters and colour matching.
- Users can choose between their own ComfyUI, Comfy Cloud and several API-provider connections instead of requiring one fixed backend.
- PSD and OpenRaster exports preserve layers, and the editor includes MCP support for compatible assistant clients.
- The core application is free software under the GPL-3.0 license.
Cons
- The project is in the 0.1.x stage and explicitly warns that users should expect rough edges.
- The Windows GitHub installer is not code-signed yet, so Windows SmartScreen may show a warning on first launch.
- Linux packages are available, but the project notes they have not been run by the author; macOS is planned rather than released.
FAQ
What is Scumble?
Scumble is a free, GPL-3.0 desktop image editor for AI-assisted inpainting, editing, upscaling and generation. It returns generated results as editable layers rather than immediately flattening them into the image.
What do I need to use Scumble's AI features?
No account or Scumble-hosted server is required. You can connect a local or remote ComfyUI, use Comfy Cloud, or enter an API key for supported providers. The editor can also open, paint, select, layer, filter, save and export without a rendering service, although AI generation requires a configured rendering destination.
Which platforms does Scumble support?
On Windows, the project provides a Microsoft Store version, a GitHub installer and a portable ZIP for 64-bit Windows 10 and 11. Releases also include Linux AppImage and .deb packages, but the project notes that these builds have not been run by the author. macOS is planned but not released.
What file formats and outputs does Scumble support?
Generated content can be exported in PSD with layers and masks or in OpenRaster with layers. Both formats can be opened again with their layers, while PNG is also supported for opening and saving.
Can an AI assistant automate Scumble?
Yes. Scumble exposes its features as commands through an MCP server. Its documentation says MCP clients such as Claude Code and Claude Desktop can use 105 tools to open, select, generate, match and export in the same window and undo context.
Quick Facts
- Category
- AI image editor
- License
- GPL-3.0 open-source software
- Primary platform
- Windows 10/11, 64-bit
- Additional platform builds
- Linux AppImage and .deb; macOS planned
- Rendering options
- Local or remote ComfyUI, Comfy Cloud, or provider API keys
- Layered outputs
- PSD and OpenRaster with layers; PNG supported
Альтернативы Scumble
Pixelcut
Pixelcut is an AI photo editor for creating and editing product images, social content, and video assets. It offers free and paid plans, plus mobile app and browser access for individual users and teams.
Upscale.media
Upscale.media is an AI image upscaling and enhancement tool from Pixelbin that enlarges low-resolution or pixelated images and improves clarity in the browser. It offers free individual image use, paid credit plans, and team-oriented workflows such as bulk processing and API integration.
Free Online AI Image Enlarger
PhotoAid’s AI Image Enlarger is a free browser-based tool for upscaling images without a standard blur-heavy resize. It supports JPEG, WEBP, and PNG uploads and is described as working best for photos with faces.
PXZ AI
Все-в-одном AI платформа, которая объединяет инструменты для изображения, видео, голоса, письма и чата для повышения креативности и сотрудничества.
Petalica Paint
Petalica Paint is a web service for automatically coloring line art and tracing line drawings from photographed images. The home page says it can be used under the stated terms for commercial and noncommercial work, but it is scheduled to end on 2025-07-31 at 16:00 JST.
PixelDrive
PixelDrive is a bulk image generation API for creating many on-brand images from one template through a browser editor, REST API, CSV batch, or MCP. It suits teams that need repeatable image output with cacheable renders and usage-based pricing.