Skip to main content
OpenComic AI Bin includes 30+ upscaling models that increase image resolution while preserving or enhancing quality. Each model has different characteristics, performance profiles, and optimal use cases.

​
Model categories

Upscaling models are organized by their underlying architecture and purpose:

​
RealCUGAN models

Advanced upscaling with noise reduction controls.

​
realcugan

  • Upscaler: realcugan
  • Scales: 2x, 3x, 4x
  • Noise levels: 0, 3
  • Latency: 2.96s
  • Source: Moebytes/waifu2x
RealCUGAN offers conservative and denoise options at each scale level. Excellent for general-purpose anime and manga upscaling.

​
RealESR models

Real-world super-resolution models optimized for anime and video content.

​
realesr-animevideov3

Optimized for anime video frames. One of the fastest models in the library.

​
RealESRGAN models

General-purpose super-resolution models with anime-specific variants.

​
realesrgan-x4plus

General-purpose model for real photos and artwork. Higher latency but excellent quality.

​
realesrgan-x4plus-anime

Anime-specialized variant with faster processing than the general model.

​
realesrnet-x4plus

RealESRNet variant focusing on natural image restoration.

​
RealESRGAN_General_WDN_x4_v3

General-purpose model with weak denoising.

​
RealESRGAN_General_x4_v3

Fastest general-purpose RealESRGAN variant.

​
Waifu2x models

Classic anime upscaling models with extensive scale options.

​
waifu2x-models-cunet

  • Upscaler: waifu2x
  • Scales: 2x, 4x, 8x, 16x, 32x
  • Noise levels: 0, 1, 2, 3
  • Latency: 5.2s (Medium)
  • Source: Moebytes/waifu2x
CUnet architecture with extensive scale support and noise reduction options.

​
waifu2x-models-upconv

  • Upscaler: waifu2x
  • Scales: 2x, 4x, 8x, 16x, 32x
  • Noise levels: 0, 1, 2, 3
  • Latency: 2.61s (Fast)
  • Source: Moebytes/waifu2x
UpConv architecture optimized for anime-style art.

​
Specialized upscaling models

​
4x-WTP-ColorDS

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.53s (Slow)
  • Source: OpenModelDB
ColorDS model for preserving color accuracy.

​
remacri-4x

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.84s (Slow)
  • Source: upscayl/upscayl
Remacri model for high-quality upscaling.

​
ultramix-balanced-4x

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 10.0s (Slow)
  • Source: upscayl/upscayl
Balanced quality/detail model.

​
ultrasharp-4x

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.73s (Slow)
  • Source: upscayl/upscayl
Focuses on edge sharpness.

​
4xInt-RemAnime

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.46s (Slow)
  • Source: Phhofm/models
Anime-focused interpolation model.

​
AI-Forever_x4plus

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.55s (Slow)
  • Source: Hugging Face
AI-Forever enhanced variant.

​
Photo and web content models

​
4xNomosWebPhoto_esrgan

  • Upscaler: upscayl
  • Scales: 2x, 3x, 4x
  • Latency: 9.77s (Slow)
  • Source: OpenModelDB
Optimized for web photos and realistic content.

​
4xHFA2k

High-fidelity artwork upscaling.

​
LSDIR models

Large-scale dataset image restoration models.

​
4xLSDIRCompactC3

Compact, fast variant of LSDIR.

​
4xLSDIRplusC

Enhanced LSDIR with improved quality.

​
Additional models

​
4x_NMKD-Siax_200k

NMKD’s Siax model trained on 200k iterations.

​
4xNomos8kSC

Nomos 8k super-compact model.

​
uniscale_restore_x4

Universal scaling with restoration focus.

​
unknown-2.0.1

Community model with unknown origin.

​
Using upscale models

import OpenComicAI from 'opencomic-ai-bin';

// Basic upscaling
await OpenComicAI.pipeline('./input.jpg', './output.jpg', [
  { model: 'realcugan', scale: 4, noise: 0 }
]);

// Upscaling with noise reduction
await OpenComicAI.pipeline('./input.jpg', './output.jpg', [
  { model: 'waifu2x-models-cunet', scale: 2, noise: 2 }
]);

// Fast upscaling
await OpenComicAI.pipeline('./input.jpg', './output.jpg', [
  { model: 'realesr-animevideov3', scale: 4 }
]);
For batch processing, consider using daemon mode to improve performance by loading models once.

​
Choosing the right model

For speed: Use realesr-animevideov3, RealESRGAN_General_x4_v3, or 4xLSDIRCompactC3 For anime/manga: Use realesrgan-x4plus-anime, realcugan, or waifu2x-models-cunet For photos: Use 4xNomosWebPhoto_esrgan or realesrgan-x4plus For extreme scales: Use waifu2x-models-cunet or waifu2x-models-upconv (up to 32x) For quality: Use realesrgan-x4plus, ultramix-balanced-4x, or ultrasharp-4x

Build docs developers (and LLMs) love

Get started for freeTalk to us