Skip to content
SPECIFICTION
Get listed

Real-CUGAN

github.com

Free plan WebWindowsmacOSLinuxAndroidSelf-hosted
Visit github.com
Spec score 8.1 #12 of 32 in AI Video Upscalers

What it is

Real-CUGAN is a free AI super-resolution model designed for anime images. It enlarges images by 2×, 3×, or 4× and offers denoising strengths, a conservative mode, and adjustable alpha. The 1× model is still in training. The English README describes five model weights for 2× and three each for 3× and 4×, covering different enhancement strengths. The Windows configuration processes both images and video, and the repository includes a video inference script. A linked NCNN implementation provides Windows, Linux, and macOS executables for Intel, AMD, Nvidia, and Apple-Silicon GPUs using Vulkan; its portable package includes binaries and models and states that it needs no CUDA or PyTorch runtime. Browser options include a demo and a Web-CPU version, though the README says the CPU demo is slow and recommends smaller images. Windows cache and tile settings can trade speed for lower GPU-memory use, including modes intended for very large images. The project also lists PyTorch and VapourSynth use paths and describes its architecture as compatible with Waifu2x-CUNet. The maker invites suggestions and requests through repository issues.

Spec sheet

Purpose
Real-CUGAN is an AI super-resolution model trained on a million-scale anime dataset and designed for anime images.Read 9 Oct 2026 · github.com
Free plan
YesRead 20 Sep 2026 · github.com
Support
The maker invites users to leave suggestions and requests in the repository issues.Read 9 Oct 2026 · github.com
Alternative GPU support
The linked NCNN implementation offers Windows, Linux, and macOS executables for Intel, AMD, Nvidia, and Apple-Silicon GPUs using Vulkan.Read 9 Oct 2026 · github.com
Desktop app
YesRead 20 Sep 2026 · github.com
Download
The maker links Windows executable packages through Baidu Drive, GitHub Releases, and Google Drive.Read 9 Oct 2026 · github.com
Enhancement models
The English README says 2× has five model weights and 3×/4× have three model weights each for different enhancement strengths.Read 9 Oct 2026 · github.com
Image and video
The Windows configuration supports both image and video processing, and the repository includes a simple video inference script.Read 9 Oct 2026 · github.com
Integration
The project describes its model architecture as compatible with Waifu2x-CUNet, and it lists PyTorch and VapourSynth use paths.Read 9 Oct 2026 · github.com
Memory limits
The Windows package provides cache and tile settings to trade processing speed for lower GPU-memory use, including modes intended for very large images.Read 9 Oct 2026 · github.com
Portable NCNN package
The NCNN package includes binaries and models and states that it needs no CUDA or PyTorch runtime.Read 9 Oct 2026 · github.com
Security and compliance
The maker README and repository page provide no security or compliance claims.Read 9 Oct 2026 · github.com
Training code
The maker README says the training code is not public.Read 9 Oct 2026 · github.com
Upscaling
It supports 2×, 3×, and 4× upscaling; the 1× model is described as still in training.Read 9 Oct 2026 · github.com
Web option
The maker README links a browser demo and a Web-CPU version; the README notes that the CPU demo is slow enough that smaller images are recommended.Read 9 Oct 2026 · github.com
Windows requirements
The package was tested on Windows 10 64-bit; Nvidia use requires 1.5 GB of video memory, with different CUDA minimums for light and heavy versions.Read 9 Oct 2026 · github.com

Plans and prices

Real-CUGAN Free Open-source model and downloadable packages; no paid tiers or usage limits stated Read 9 Oct 2026 · github.com

Every Real-CUGAN plan and price →

Strong points and limits

Strong points

  • Supports 2×, 3×, and 4× upscaling
  • Includes multiple enhancement strengths
  • Windows configuration supports images and video
  • Portable NCNN package needs no CUDA or PyTorch runtime
  • Cache and tile settings reduce GPU-memory use

Limits

  • 1× model is still in training
  • Windows Nvidia use requires 1.5 GB of video memory
  • CPU demo is slow; smaller images are recommended
  • Training code is not public

Who it suits

Real-CUGAN suits people seeking anime-focused image enlargement, including users who want to process video or use GPU-specific implementations. The browser demo offers another route, though the CPU version is described as slow.

Verdict

Real-CUGAN provides several anime upscaling strengths and paths for desktop, GPU, and browser use. Consider the stated Windows memory requirements and slower CPU demo when choosing how to run it.

Where Specifiction places it

Alternatives to Real-CUGANAll alternatives →