Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ollm/opencomic-ai-training/llms.txt

Use this file to discover all available pages before exploring further.

OpenComic AI Training is an ethical dataset generation pipeline that procedurally synthesizes paired clean and degraded comic image datasets inside Krita. It powers the OpenComic AI models for artifact removal, halftone descreening, and 2×/3×/4× upscaling — all trained exclusively on procedurally generated, copyright-free images.

Quickstart

Install dependencies, build the project, and generate your first dataset in minutes.

Requirements

Everything you need: Linux, Krita 5.3+, the kra-remote plugin, and Node.js.

Pipeline Overview

Understand how Krita, Node.js, and YAML options work together to produce training data.

Configuration Reference

Full schema for options files — seeds, drawings, degradations, and output paths.

Model Reference

Browse all pre-trained ESRGAN, Lite, and Compact model weights in NCNN format.

CLI Reference

Every flag for the generate command and the dataset validation utility.

How it works

1

Install and build

Clone the repository, run npm install, then npm run prepare to compile TypeScript and bundle the output.
2

Choose or write an options file

Pick one of the pre-built YAML presets in options/ — such as opencomic-ai-upscale-2x.yml — or create your own by composing common building blocks.
3

Run the generator

Point the generator at your options file and your Krita AppImage. It launches Krita, draws procedural comic art, applies synthetic degradations, and saves paired clean/ and degraded/ images.
4

Train your model

Feed the generated dataset to any paired-image training framework. The repository includes ready-to-use traiNNer-redux training configs in options/train/.

Dataset tasks

Artifact Removal

Remove JPEG, WebP, AVIF, and JXL compression artifacts from scanned or re-encoded comic images.

Descreen

Eliminate halftone dot patterns from printed comics scanned at various resolutions and angles.

Upscale

Super-resolve comic pages at 2×, 3×, or 4× scale while preserving fine linework detail.

Build docs developers (and LLMs) love