Skip to main content

Documentation Index

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

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

Artifact removal models reduce the visual damage introduced by lossy JPEG compression: block boundaries, ringing around hard edges, colour banding, and the muddy loss of fine detail that accumulates through repeated saves. Unlike upscaling models, every model in this category operates at 1× scale — the output image has exactly the same pixel dimensions as the input, with only the compression noise cleaned up. All models in this category use the upscayl ESRGAN-compatible backend, and their .bin/.param weight files are located under <modelsPath>/artifact-removal/models/.
These models do not change image resolution. The output dimensions will always match the input. To both clean and enlarge an image, run artifact removal first, then pass the result through an upscaling model.

​
All Artifact Removal Models

Model IDDisplay NameNotes
opencomic-ai-artifact-removalOpenComic AI Artifact RemovalStandard tier; strongest noise suppression
opencomic-ai-artifact-removal-liteOpenComic AI Artifact Removal LiteLite tier; faster with slightly reduced suppression
opencomic-ai-artifact-removal-compactOpenComic AI Artifact Removal CompactCompact tier; fastest and smallest — ideal for batch processing
1x_NMKD-Jaywreck3-Lite_320kNMKD Jaywreck3 LiteAggressive JPEG cleaning; NMKD.de
1x_NMKD-Jaywreck3-Soft-Lite_320kNMKD Jaywreck3 Soft LiteGentler variant of Jaywreck3; preserves more texture
1x-SaiyaJin-DeJpegSaiyaJin DeJpegTargeted JPEG block removal; OpenModelDB
1x_JPEGDestroyerV2_96000GJPEG Destroyer V2Heavy-handed artefact crushing; Hugging Face

​
OpenComic AI Three-Tier System

The three custom OpenComic AI artifact removal models are trained on the same dataset but compiled at different network capacities, giving you a straightforward speed-vs-quality trade-off:
TierModel IDCharacteristic
Standardopencomic-ai-artifact-removalLargest network; highest fidelity artefact suppression — use when quality is the priority
Liteopencomic-ai-artifact-removal-liteReduced network size; noticeably faster while retaining most quality gains
Compactopencomic-ai-artifact-removal-compactSmallest and fastest; suited for real-time previews or large batch jobs where throughput matters
Start with Lite for most workflows. Upgrade to Standard only if the output still shows visible JPEG blocking; drop to Compact when processing hundreds of images or when working under memory constraints.

​
Community Models

The four community models cover different points on the aggression spectrum:
  • NMKD Jaywreck3 Lite and its Soft Lite variant are general-purpose JPEG cleaners. The soft variant applies less aggressive smoothing and is a better choice when you want to remove artefacts without softening fine linework.
  • SaiyaJin DeJpeg targets the characteristic 8×8 DCT block pattern directly, making it effective at low-to-medium quality JPEG sources.
  • JPEG Destroyer V2 applies very strong suppression and is best reserved for heavily compressed scans where blocking is the dominant problem.

​
Usage

import OpenComicAIModels from 'opencomic-ai-models';
import { execFile } from 'node:child_process';
import path from 'node:path';

const modelsPath = OpenComicAIModels.path;

execFile('upscayl-bin', [
  '-i', 'input.jpg',
  '-o', 'output.png',
  '-m', path.join(modelsPath, 'artifact-removal', 'models'),
  '-n', 'opencomic-ai-artifact-removal-lite',
  '-s', '1',
]);
Shell
upscayl-bin \
  -i input.jpg \
  -o output.png \
  -m /path/to/models/artifact-removal/models \
  -n opencomic-ai-artifact-removal-lite \
  -s 1

Build docs developers (and LLMs) love

Get started for freeTalk to us