# OpenLayer

> OpenLayer is a free, open-source (MIT) ComfyUI plugin for Adobe Photoshop. It is a UXP panel that connects Photoshop to a ComfyUI server running on your own computer, so you can generate and edit images with local AI models and get every result back as a real, editable Photoshop layer. No cloud, no subscription, no account, no credits, and nothing is uploaded.

OpenLayer is for anyone who uses Photoshop and wants AI image generation without Adobe Firefly's generative credits or a hosted service: photographers, designers, digital artists, retouchers, and ComfyUI users who are tired of moving images between ComfyUI and Photoshop by hand. It works as a free, local alternative to Photoshop Generative Fill.

It is alpha software. It is tested on Windows 11 with Photoshop 2025; macOS is untested.

Requirements: Photoshop 2024 or newer, a local ComfyUI server, and a GPU with 8 GB VRAM or more (12 GB recommended).

Installation: download openlayer-latest.ccx from the GitHub releases and double-click it; the Adobe Creative Cloud desktop app installs it. Building from source and loading it with the Adobe UXP Developer Tool is the alternative.

## Tools

- Text to Image: generate a new layer from a prompt, including transparent cut-out layers
- Image to Image: use the active layer as visual input
- Edit Image: change a layer or a selection by describing the change in words
- Sketch to Image: guide generation with your own line art (ControlNet)
- Inpaint: repaint a Photoshop selection in place
- Outpaint: extend the canvas beyond its edges
- Upscale: enhance a layer
- Remove Background: cut the subject onto its own layer with real alpha
- Layer Maps: depth, line-art and normal maps from a layer
- Prompt from Layer: describe a layer as prompt text
- Unflatten (experimental): split one flat image into separate layers with transparency, using Qwen-Image-Layered
- Live Painting: paint and watch the model respond in real time
- Style Reference and Multi-Reference: borrow a reference layer's mood, or compose one image from several layers
- Agent Bridge (MCP): lets AI agents drive the panel's tools

## Supported models

FLUX.2 Klein 4B (Apache-2.0), Qwen-Image 2.1, Qwen-Image-Layered, Krea-2 Turbo, Z-Image Turbo, Flux Fill, and Stable Diffusion 1.5 checkpoints with your own LoRAs. Qwen-Image 2.1 weights are licensed for research and evaluation only, and Flux Fill is non-commercial, so those presets are opt-in and labelled. Check each model's licence before commercial use.

## Links

- [GitHub repository](https://github.com/MehranMarxian/OpenLayer): source code, README, and issue tracker
- [Download (.ccx)](https://github.com/MehranMarxian/OpenLayer/releases/latest/download/openlayer-latest.ccx)
- [Releases and changelog](https://github.com/MehranMarxian/OpenLayer/releases)
- [Getting started](https://github.com/MehranMarxian/OpenLayer/blob/main/docs/getting-started.md)
- [Known limitations](https://github.com/MehranMarxian/OpenLayer/blob/main/docs/known-limitations.md)
- [Discussions](https://github.com/MehranMarxian/OpenLayer/discussions)
- [Website](https://mehran-ahmadi.com/OpenLayer/)

## Author

Built by Mehran Ahmadi, a photographer and digital artist. GitHub: MehranMarxian.