🇺🇸English

Upscale Image

Prerequisites

• uv (Python package manager)
• GPU with Vulkan support (Metal on macOS, Vulkan on Linux/Windows)

The realesrgan-ncnn-vulkan binary and model files are automatically downloaded to ~/.cache/realesrgan-ncnn-vulkan/ on first run.

Usage

Run from the skill folder (skills/upscale-image/):

Upscale a single image (4x by default)

uv run --project runtime runtime/upscale.py path/to/image.png

Specify scale factor

uv run --project runtime runtime/upscale.py image.png -s 2

Choose a model

uv run --project runtime runtime/upscale.py image.png -m realesrgan-x4plus

Available models:

• realesr-animevideov3 (default) — fast, lightweight, good for most uses
• realesrgan-x4plus — highest quality general-purpose (slower)
• realesrgan-x4plus-anime — optimized for anime/illustrations

Specify output directory

uv run --project runtime runtime/upscale.py image.png -o path/to/output/

Batch processing

uv run --project runtime runtime/upscale.py img1.png img2.jpg img3.webp

Output

• Output files are saved as PNG.
• Output filenames follow the pattern <original_name>_upscaled.png.
• Default output directory: output/ inside the skill folder.

Limitations

• realesrgan-x4plus is significantly slower than realesr-animevideov3 on large images.
• Supported input formats: PNG, JPEG, WebP.
• Binary source: Real-ESRGAN-ncnn-vulkan
• Model source: Real-ESRGAN releases

Weekly Installs

1

Repository

anyt-io/notebook

GitHub Stars

9

First Seen

1 day ago

Security Audits

Gen Agent Trust HubWarnSocketWarnSnykWarn

Installed on

mcpjam1

github-copilot1

junie1

windsurf1

zencoder1

crush1