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.

Upscaling models increase image resolution by a fixed scale factor (2×, 3×, or 4×) while reconstructing fine detail that would otherwise be lost from simple interpolation. The opencomic-ai-models package ships all model weights as NCNN .bin and .param file pairs, organized into three subdirectories that map directly to their required upscaler binary: upscayl (ESRGAN-based), realcugan (Real-CUGAN), and waifu2x (Waifu2x). When invoking an upscaler binary you pass the model ID — the filename stem, without extension — as the model name argument, along with the path returned by OpenComicAIModels.path pointing to the correct subdirectory.

​
OpenComic AI Models

Custom models trained specifically for comic and manga content, available in three tiers and multiple scale factors. All use the upscayl backend. Model files are located under <modelsPath>/upscale/models/.
Model IDDisplay NameScaleNotes
opencomic-ai-upscale-4xOpenComic AI Upscale 4x4×Standard quality; best detail reconstruction
opencomic-ai-upscale-4x-liteOpenComic AI Upscale 4x Lite4×Faster inference, slightly reduced detail
opencomic-ai-upscale-4x-compactOpenComic AI Upscale 4x Compact4×Fastest, smallest footprint
opencomic-ai-upscale-3x-liteOpenComic AI Upscale 3x Lite3×Lite tier for 3× output
opencomic-ai-upscale-2xOpenComic AI Upscale 2x2×Standard quality for 2× output
opencomic-ai-upscale-2x-liteOpenComic AI Upscale 2x Lite2×Faster 2× inference
opencomic-ai-upscale-2x-compactOpenComic AI Upscale 2x Compact2×Fastest 2× option
For the best balance of speed and quality on comic or manga pages, start with opencomic-ai-upscale-4x-lite. Only move to the standard tier if you need maximum sharpness on fine linework.

​
RealESRGAN / upscayl

Official RealESRGAN and RealESRNet models from xinntao/Real-ESRGAN, converted to NCNN format and used with the upscayl backend. Files are located under <modelsPath>/upscale/models/.
Model IDDisplay NameScaleNotes
realesrgan-x4plusRealESRGAN x4 Plus4×General-purpose photo upscaling
realesrgan-x4plus-animeRealESRGAN x4 Plus Anime4×Tuned for flat-color anime/manga art
realesrnet-x4plusRealESRNet x4 Plus4×Lighter network; faster than x4plus
realesr-animevideov3-x2RealESR AnimeVideo v3 2×2×Anime video model, 2× scale
realesr-animevideov3-x3RealESR AnimeVideo v3 3×3×Anime video model, 3× scale
realesr-animevideov3-x4RealESR AnimeVideo v3 4×4×Anime video model, 4× scale
realesrgan-x4plus-anime is the go-to choice for comic and manga content. Its training distribution closely matches screentone-free line art and flat colour fills, producing crisp edges with minimal haloing.

​
RealCUGAN

Real-CUGAN is a high-quality upscaler for anime-style images. It uses its own dedicated realcugan binary. The model files are located under <modelsPath>/upscale/realcugan/.
Model IDDisplay NameScaleNotes
realcuganRealCUGAN2×/3×/4×High-quality anime upscaling; requires the realcugan binary

​
Waifu2x

Classic Waifu2x models, served via the waifu2x binary. Files are located under <modelsPath>/upscale/waifu2x/.
Model IDDisplay NameScaleNotes
waifu2x-models-cunetWaifu2x CUnet1×/2×Higher quality; slower
waifu2x-models-upconvWaifu2x UpConv1×/2×Faster, slightly lower quality

​
Community Models

A curated set of open community models bundled for convenience, all using the upscayl backend. Files are located under <modelsPath>/upscale/models/.

​
General / Photo

Model IDDisplay NameScaleNotes
4x-WTP-ColorDSWTP ColorDS4×Colour-detail preserving; OpenModelDB
4xNomosWebPhoto_esrganNomos Web Photo ESRGAN4×Optimised for web-sourced photos; OpenModelDB
AI-Forever_x4plusAI-Forever x4plus4×RealESRGAN fine-tune; Hugging Face
RealESRGAN_General_x4_v3RealESRGAN General v34×Broad-purpose general upscaler
RealESRGAN_General_WDN_x4_v3RealESRGAN General WDN v34×Adds wavelet denoising pass
uniscale_restore_x4Uniscale Restore x44×Restoration-focused upscaler
unknown-2.0.1Unknown 2.0.14×Community model from upscayl/custom-models

​
Anime / Illustration

Model IDDisplay NameScaleNotes
4xInt-RemAnimeInt-RemAnime4×Anime upscaling with artefact reduction; Phhofm/models
2x-AnimeSharpV4_RCAN_fp16_op17AnimeSharp V4 RCAN2×RCAN architecture; sharp anime output
4x_NMKD-Siax_200kNMKD Siax4×Anime-style; high-frequency detail
4xHFA2kHFA2k4×Anime line preservation
4xNomos8kSCNomos 8k SC4×Screened/scanned content focus
4xLSDIRCompactC3LSDIR Compact C34×Compact LSDIR architecture
4xLSDIRplusCLSDIR Plus C4×Full LSDIR Plus architecture

​
Upscayl Bundled

Model IDDisplay NameScaleNotes
remacri-4xRemacri4×Soft, natural texture rendering
ultramix-balanced-4xUltramix Balanced4×Balanced sharpness and smoothness
ultrasharp-4xUltrasharp4×Maximum sharpness; may emphasise noise

​
Usage

Retrieve the models base path from the package, then pass it together with the desired model ID to the appropriate upscaler binary. 
import OpenComicAIModels from 'opencomic-ai-models';
import { execFile } from 'node:child_process';
import path from 'node:path';

const modelsPath = OpenComicAIModels.path;

// upscayl-based model (ESRGAN)
execFile('upscayl-bin', [
  '-i', 'input.png',
  '-o', 'output.png',
  '-m', path.join(modelsPath, 'upscale', 'models'),
  '-n', 'realesrgan-x4plus-anime',
  '-s', '4',
]);

upscayl-bin \
  -i input.png \
  -o output.png \
  -m /path/to/models/upscale/models \
  -n realesrgan-x4plus-anime \
  -s 4

Build docs developers (and LLMs) love

Get started for freeTalk to us