Image Upscaling

This wraps the inference.sh CLI to give Claude direct access to several image upscaling models, with Topaz as the main option after Real-ESRGAN and others were apparently cycled out. You can feed it low resolution images and get back higher quality versions, which is useful for cleaning up AI generated art before printing or restoring old photos. The skill focuses on a generate-then-upscale workflow, pairing nicely with FLUX or other image generation models. It's straightforward if you're already in the inference.sh ecosystem, though you'll need to install and authenticate with their belt CLI first. The documentation is solid and the skill page shows actual command examples rather than just talking about possibilities.

install copies · 90 days. Times the install command was copied on this site in the last 90 days, bots filtered.: 9
skills.sh installs: 1.1K
GitHub stars: 804
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add inference-sh/skills --skill image-upscaling --agent claude-code

Installs into .claude/skills of the current project.

Files

Install the belt CLI skill: npx skills add belt-sh/cli

Image Upscaling

Upscale and enhance images via inference.sh CLI.

Image Upscaling

Quick Start

Requires inference.sh CLI (belt). Install instructions

belt login

belt app run infsh/real-esrgan --input '{"input_file": "https://your-image.jpg"}'

Available Upscalers

Model	App ID	Best For
Topaz Image Upscaler	`falai/topaz-image-upscaler`	Professional quality, any image

Examples

Upscale Any Image

belt app run falai/topaz-image-upscaler --input '{"image": "https://low-res-image.jpg"}'

Workflow: Generate and Upscale

# 1. Generate image with FLUX Klein (fast)
belt app run falai/flux-2-klein-lora --input '{"prompt": "landscape painting"}' > image.json

# 2. Upscale the result
belt app run falai/topaz-image-upscaler --input '{"image": "<url-from-step-1>"}'