The anti-filter: redrawing snapshots as sparse diary pages

What it does photo-revival is a Codex skill—a curated set of system prompts and constraints—that turns everyday snapshots into 3:4 vertical illustrations. Instead of slapping on a filter, it treats the source image as reference material and redraws the scene as a small pencil or watercolor sketch floating on an otherwise blank white page. The goal is a field-note aesthetic, not a polished photo edit.

The interesting bit The README is essentially a manifesto for negative space. It obsessively quantifies the layout—the subject must occupy only 10–16% of the canvas (hard ceiling at 18%), color must stay trapped inside that zone, and 80–88% of the paper must remain untouched white—because image models naturally want to fill the frame. It is prompt engineering wielded as graphic-design enforcement.

Key highlights

• Rejects style transfer in favor of genuine redrawing; the output is an illustration, not a filtered photo
• Enforces a strict 3:4 vertical format with 80–88% blank white paper and tiny integrated handwritten captions
• Ships with corrective prompt templates to fight model drift: subject bloat, color bleed, duplicated text, and poster-style typography
• Openly lists its own boundaries: no portrait retouching, commercial posters, dense infographics, or pixel-accurate restoration
• Runs as a skill inside OpenAI Codex, invoked with $photo-revival

Caveats

• The FAQ openly admits the model often ignores spatial constraints, requiring users to paste stricter reinforcement prompts and regenerate
• Handwritten text sometimes repeats or garbles, so the workflow expects manual cleanup passes
• It is not a standalone tool; you need the Codex environment to use it

Verdict Worth a look if you want your alleyway snapshots and lunch photos to feel like archival diary entries. Avoid if you need a general-purpose image pipeline or a fire-and-forget API.