Krita ComfyUI

Krita ComfyUI is a lightweight plugin for Krita 5.2+ that lets you run ComfyUI workflows directly from Krita. The plugin focuses on reliable communication with the ComfyUI server; workflow design and configuration are handled by the user.

⚠️ Compatibility: Krita 5.2 or newer. ⚙️ GPU: To generate images locally you need a GPU with at least 6 GB VRAM (NVIDIA, AMD, or Apple Silicon).

---

Features

Feature	Description
Generate	Run ComfyUI workflows to produce images.
Prompt	Text‑to‑image from a simple prompt dialog.
Inpainting	Use Krita selections for generative fill, expansion or object removal.
History	Preview the images generated during the current Krita session.
Cloud execution	Run workflows on ComfyUI Cloud by providing an API key.
Generation timeout	Configure how long a generation can run before being aborted.
Configure	Simple UI to connect to your ComfyUI server and select workflows.
UI Style	Built‑in style presets for a streamlined interface.

---

Getting Started

1. Install Krita

   • Download the latest version from krita.org.
   • Minimum version: 5.2.0.

2. ComfyUI Server The plugin does not install or run ComfyUI; it must already be running before opening Krita. You can use the official version available at ComfyUI.

3. Download & Install the Plug‑in

   Step	Action
   1	Grab the latest release ZIP: https://github.com/dacert/krita-comfyui/releases/latest
   2	In Krita, go to Tools ▸ Scripts ▸ Import Python Plugin from File… and select the ZIP.
   3	Restart Krita.

Initial Setup

👉 If you run ComfyUI locally, start the server first before opening Krita so the plugin can detect it automatically.

1. Show the ComfyUI Docker: Settings ▸ Dockers ▸ Krita ComfyUI.

2. In the Docker window, click the settings icon and fill in the tabs:

   • General
     • ComfyUI URL – e.g. http://localhost:8000/ for a local server, or your ComfyUI Cloud endpoint for cloud execution.
     • API Key (optional, for ComfyUI Cloud) – leave empty for a local server. When using ComfyUI Cloud, paste the API key issued by the cloud service; it is sent as the X-API-Key header (HTTP) and as a token query parameter on the WebSocket connection. The key is stored obfuscated in ~/.krita_comfyui/secrets.json, separate from the main configuration.
     • Generation timeout (minutes) – maximum time a single generation is allowed to run (1–60, default 5). Increase it for slow workflows such as high‑resolution or multi‑pass pipelines.
     • Enable clipspace uploads – when enabled (default) the plugin uploads the full clipspace chain (mask, paint, painted, painted‑mask) before running inpainting workflows. Disable to upload only the base image and the selection‑reduced mask, which is what non‑clipspace inpaint pipelines actually consume.
   • Workflow – Select a workflow and adjust its inputs.

   

Running on ComfyUI Cloud

ComfyUI Cloud lets you run workflows on hosted GPUs instead of your own hardware. To use it from the plugin:

1. Subscribe to ComfyUI Cloud and copy your API key from the dashboard.
2. In Settings ▸ General, set:
   • ComfyUI URL to https://cloud.comfy.org.
   • API Key to the value copied in step 1.
3. Pick a workflow in the Workflow tab and click Generate as usual.

The key is never sent anywhere other than to the configured ComfyUI server.

💡 Tip: For extra security, set the KRITA_COMFYUI_API_KEY environment variable instead — the key will never touch the filesystem.

Workflow Configuration

To have the plugin list available workflows, the ComfyUI server must be online and contain them.

⚠️ Critical Requirement: Workflows must include a SaveImageWebsocket output node. This custom node is required for the plugin to receive generated images from ComfyUI. Without it, execution will fail with the error: "No 'SaveImageWebsocket' output node found."

1. In the Workflow tab, select a workflow (e.g., image_qwen_image_edit_2509.json).

2. Use Configure Inputs... to open the configuration dialog or Remove Configuration to delete it.

   

3. Link parameters to "Krita Inputs" node: prompt, seed, image loading nodes, etc.

   

⚠️ Important: This graph configuration dialog is read-only — it is not intended for editing workflows. Its only purpose is to link the Krita input node with the other nodes, establishing the communication interface between the plugin and the workflows.

Using the Plugin

1. Basic Generation

• Select a workflow from the drop‑down menu.
• Enter a prompt (e.g., “a cat”).
• Click Generate to get the image.

2. Inpainting

1. Open an image and select the area to modify with any Krita selection tool.

   🧪 Experimental: in Settings ▸ General, the Enable clipspace uploads option controls what gets uploaded for inpaint workflows. Leave it on to send the full clipspace chain (mask, paint, painted, painted‑mask). Turn it off to send only the base image and the mask reduced by the selection — usually what plain inpaint pipelines need. This setting is experimental and does not affect the final image quality; it only changes which intermediate files are uploaded.

2. Choose the Qwen Image Edit – Inpaint.json workflow.

3. Enter a prompt (e.g., “replace carrots with broccoli”).

4. Click Generate and, when satisfied, click Apply to insert the generated image.

   

3. Image Editing

• Select the area to edit.

• Choose the Qwen Image Edit.json workflow (or any that accepts an image input).

• Enter a descriptive prompt (e.g., “add a pink ribbon to the rabbit’s neck”).

• Generate and apply.

  

4. History

The plugin temporarily stores all generations made during the session. You can browse the history to preview previous results; it is lost when Krita closes.

Supported Platforms & Hardware

OS	GPU support
Windows, Linux, macOS	• NVIDIA – CUDA (Win/Linux) • AMD – DirectML (Win; limited), ROCm (Linux) • Apple Silicon – MPS (macOS 14+) • CPU – Very slow • XPU – Supported but may be slower

Tip: A powerful GPU (≥6 GB VRAM) will drastically improve generation speed.

Known Issues

• Workflows containing node-groups are not yet supported.
• If you want to use a pre‑configured workflow, start ComfyUI before opening Krita; otherwise the plug‑in won’t load it correctly.

---

Troubleshooting

Problem	Possible Cause	Solution
Server URL not shown	ComfyUI server isn’t running or URL is incorrect	Start the server and verify the URL.
Workflows don’t list	Connection failed to the server	Check that the firewall allows connections to localhost:8000 (or the configured port).
Plugin fails to generate images	Insufficient GPU or outdated drivers	Update your graphics card drivers and check available VRAM.
SaveImageWebsocket node not found	Workflow missing required output node	Add a SaveImageWebsocket node to your ComfyUI workflow and reconnect it to the output image pipe.
Cloud requests return 401/403	Missing or invalid API key	Verify the API key in Settings ▸ General ▸ API Key.
Generation aborts with a timeout	timeout_minutes too low for the workflow	Raise Generation timeout in Settings ▸ General.
API key looks wrong / leaked	The key may have been compromised	Rotate the key in ComfyUI Cloud and update it in Settings ▸ General ▸ API Key.

Contribution and Support

• Contribute – We welcome contributions! Please read our contributing guide before submitting a pull request.
• Report bugs / questions – Open an issue on GitHub: https://github.com/dacert/krita-comfyui/issues. Do not use official Krita channels for plugin support.

Happy creation! 🎨