krita-ai-diffusion  by Acly

Krita plugin for generative AI image editing

created 1 year ago
8,920 stars

Top 5.8% on sourcepulse

GitHubView on GitHub
Project Summary

This plugin provides a streamlined interface for generative AI image creation and editing directly within Krita, targeting digital artists and designers. It enhances creative workflows by integrating AI generation, inpainting, and upscaling with Krita's existing tools, offering precise control and seamless integration.

How It Works

The plugin leverages ComfyUI as its backend for image generation, supporting Stable Diffusion models (1.5, XL, Flux, SD3 partial). It emphasizes precision and workflow integration, allowing users to restrict AI generation to selections, refine existing content with adjustable strength, and guide generation using various inputs like sketches, line art, and depth maps via ControlNet and IP-Adapter. This approach aims to provide artists with granular control over AI outputs within their familiar painting environment.

Quick Start & Requirements

  • Installation: Install Krita (v5.2.0+), download the plugin ZIP, and import it via Tools ▸ Scripts ▸ Import Python Plugin from File.... Restart Krita and enable the docker via Settings ‣ Dockers ‣ AI Image Generation.
  • Hardware: A powerful NVIDIA GPU with at least 6 GB VRAM is recommended for local generation. AMD GPU DirectML (Windows, 12+ GB VRAM recommended) and Apple Silicon (MPS on macOS 14+) are supported with limitations. CPU generation is possible but very slow.
  • Backend: Uses ComfyUI, which can be installed manually or via the plugin's automatic setup.
  • Resources: Local generation requires significant VRAM.
  • Docs: Plugin Installation Guide, User Guide, FAQ.

Highlighted Details

  • Supports Stable Diffusion 1.5, XL, Flux, and SD3 (partial).
  • Integrates ControlNet (Scribble, Line art, Canny, Pose, Depth, etc.) and IP-Adapter for guided generation.
  • Features include inpainting, live painting, upscaling, job queue, and history.
  • Allows customization of presets with custom checkpoints, LoRAs, and samplers.
  • Supports assigning regional text prompts via layers.

Maintenance & Community

  • Active development with a community forum for discussions and a Discord server for support.
  • Contributions are welcomed via a contributing guide.
  • Community: Discussion, Discord.

Licensing & Compatibility

  • The plugin itself appears to be under a permissive license, but it relies on ComfyUI and Stable Diffusion models, which have their own licenses. Users should verify compatibility for commercial use based on the specific models and backend components used.

Limitations & Caveats

Local generation performance is heavily dependent on user hardware, with insufficient VRAM leading to slow speeds or failures. AMD GPU support via DirectML has limited features. The plugin is an extension to Krita and may have compatibility issues with future Krita versions or other Krita plugins.

Health Check
Last commit

2 weeks ago

Responsiveness

1 day

Pull Requests (30d)
6
Issues (30d)
41
Star History
496 stars in the last 90 days

Explore Similar Projects

Feedback? Help us improve.