upscale

upscale

Upscale images and videos with a fast and easy to use cross-platform CLI built around proven open-source models.

upscale gives you a simple command surface for common enhancement workflows, whether you are restoring old footage, improving video content, or preparing visuals for higher-resolution displays.

Highlights

• One CLI toolkit for both image and video upscaling
• Multiple engines for different content styles
• Hardware acceleration for Intel, AMD and Nvidia GPUs
• Advanced Anime4K shader combining to achieve the best results
• Fully offline and self-hosted
• Cross-platform

Current implementations

[^1]: Assumes a relatively modern GPU

Installation

Arch Linux

yay -S upscale-git

Binaries (Windows, MacOS, Linux)

Download pre-made binaries for Windows, MacOS and Linux from releases.

Build from source (Linux)

Requirements:

• Go
• ffmpeg
• realesrgan-ncnn-vulkan
• Python 3.10+

Build:

go build -o upscale .

Setup Python and SwinIR environment:

./install.sh

Quick Start

General command shape:

upscale <image|video> -i <input> -o <output> [options]

Image example

upscale image -i input.jpg -o output.png -a realesrgan -m normal -s 4 -q 3

Video example

upscale video -i input.mp4 -o output.mp4 -a anime4k-video -s 4 -q 4 -p '{"encoder": "av1"}'

Common Options

• -a: Select engine/adapter
• -s: Upscale factor
• -q: Quality preset (1 low, 2 medium, 3 high, 4 ultra)
• -m: Content style (normal or anime)
• -p: JSON for engine-specific tuning
• -r: Raw progress output (useful for wrappers and scripts)

Run help for all options:

upscale image -h
upscale video -h