Skip to main content

AI Tools

Nightly

OpenComic includes AI-based image tools focused on comic and manga restoration workflows.

The currently restoration workflows are the following:

These tools are intended to improve image quality while preserving as much of the original artwork as possible.

OpenComic AI models​

OpenComic provides its own family of models under the OpenComic AI name.

These models are trained with ethically generated images created procedurally and randomly in Krita. The stated goal of the project is to preserve as much detail from the original image as possible while modifying only the necessary parts.

For more information about how these models were trained, visit the main repository.

Quality levels of OpenComic AI models​

OpenComic AI models come in three distinct quality levels: Fast, Balanced, and Quality. Each represents a different balance between speed and output quality.

  • Fast: The quickest option with the lowest output quality.
  • Balanced: A balanced trade-off between quality and speed.
  • Quality: The highest output quality but also the slowest option.

Model versions​

OpenComic AI includes multiple model versions with bug fixes and quality improvements. To see the changes between versions, check the CHANGELOG.md in the main repository.

Third-party models​

OpenComic AI tooling can also use models and upscalers from third-party projects.

The models and runtimes from projects such as:

  • Real-CUGAN.
  • Real-ESRGAN.
  • Waifu2x.
  • Upscayl and models distributed through its ecosystem.
  • Other third-party models listed in the package model tables.
warning

OpenComic documents the OpenComic AI models as ethically generated and trained. For third-party models, this documentation does not verify or guarantee their dataset origin, training methodology, or ethical status.

If you need strict provenance requirements, prefer the OpenComic AI models and review the source repositories of any third-party model before using it.

What each tool does​

Artifact removal​

Artifact removal is designed to reduce compression noise and similar defects without changing more of the image than necessary.

AI artifact removal input
AI artifact removal output
InputOutput
AI artifact removal input
AI artifact removal output
InputOutput

Descreening​

Descreening removes halftone and reduces moiré patterns that often appear in scanned manga and comics.

AI descreen input
AI descreen output
InputOutput
AI descreen input 2
AI descreen output 2
InputOutput
AI descreen input 3
AI descreen output 3
InputOutput

Upscaling​

Upscaling increases image resolution so pages can look cleaner on high-resolution displays or after enlargement.

AI upscale input
AI upscale output
InputOutput
AI upscale input
AI upscale output
InputOutput

Extract document images​

When you use AI Tools with PDF or EPUB files, OpenComic needs to work with page images instead of the original document container.

For that reason, Extract document images is required for AI processing on PDF and EPUB files.

This extraction step extracts images from inside PDF or EPUB that the AI tools can process.

Licensing​

In the public OpenComic AI repositories:

  • Source code is published under the MIT license.
  • OpenComic AI model weights are published under CC BY 4.0.

For third-party models, their own licenses and terms apply.