Upload a badly exposed or blurry photo and get four enhanced versions ready to share or print.
Restore an old, faded, or scratched family photo while keeping it looking natural.
Enhance a landscape or travel photo with more dramatic lighting and composition.
Improve a museum, product, or document photo while keeping text, logos, and details accurate.
| justlovemaki/lumarescue | abgcto/hey-claude | amap-ml/blockpilot | |
|---|---|---|---|
| Stars | 62 | 62 | 62 |
| Language | — | Swift | Python |
| Setup difficulty | easy | easy | moderate |
| Complexity | 2/5 | 2/5 | 3/5 |
| Audience | general | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Runs as a skill inside an AI desktop tool like Codex or Doubao rather than as standalone software.
LumaRescue is a photo editing skill built to run inside AI desktop tools such as Codex or Doubao's desktop app. Its purpose is to take poorly shot photos, ones that are badly exposed, off in color, blurry, low contrast, or badly framed, and turn them into images that are ready to share, print, or build on further. The README describes it as working on portraits, landscapes, architecture, museum pieces, products, documents, and old or faded photos, including scans, scratches, and damaged areas. The core idea is a split between what must stay fixed and what can change. A person's identity, facial structure, expression, pose, clothing, readable text, logos, and factual details of buildings or artifacts are kept exactly as they are. Lighting, color, sky, background clutter, cropping, and overall composition are treated as flexible and can be reworked to make the photo more striking. The README states that lighting is handled before color and filters, aiming for a photo with clear light and shadow instead of one that is simply brighter or more saturated. By default the skill delivers four labeled versions of each photo: a natural programmatic edit, a more stylized programmatic edit, a natural AI-assisted edit, and a bolder AI-assisted edit that allows bigger changes to lighting, mood, sky, and framing. It also returns the text prompts used to generate the AI versions, so a user can reuse or adjust them. The original photo is not included in the output by default, since the README treats it as internal reference only. To use it, someone uploads a photo and gives an instruction such as asking the skill to rescue the uploaded photo, in either English or Chinese, and can add specific goals like restoring an old photo naturally or brightening a landscape while keeping the real scenery. The project includes reference guides on photo diagnosis and photographic style that guide these automatic decisions. The README notes documents, artifacts, and museum photos should stay accurate, while travel and landscape photos are given more room for dramatic enhancement.
An AI skill for desktop coding assistants that repairs and beautifies bad photos into four labeled versions while preserving a subject's identity and factual details.
Not sufficiently specified in the provided material.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.