Skip to content

project.md

Latest commit

 

History

History
81 lines (67 loc) · 3.47 KB

project.md

File metadata and controls

81 lines (67 loc) · 3.47 KB

Core concept (one sentence)

An app that helps amateur photographers improve by analyzing their photos for common problems, showing visual guidance drawn over the actual photo (crop lines, arrows, red “sharpie” notes), and letting users apply constrained AI edits (lighting, background cleanup, props) while learning why each change helps.

User flows (mobile-first)

  1. Camera / upload → choose mode: Social / Pro / Practice (controls how aggressive suggestions can be).
  2. Auto-analysis: composition, exposure, sharpness, white balance, background clutter, subject isolation, color harmony.
  3. Visual feedback: overlay on the photo (rule-of-thirds grid, arrows, circled problem areas, suggested crop box).
  4. Edit controls: toggles for allowed edits (lighting only / crop only / background blur / remove small objects). Preview slider for before/after.
  5. Learn & repeat: short micro-lesson explaining each suggestion and a simple re-shoot prompt (e.g., “move subject left 1 step”).
  6. Save versions and export (web/social sizes).

Feature ideas (prioritized)

Core / MVP

  • Auto-analysis for composition (rule of thirds, headroom), exposure (blown highlights/shadows), blur detection, and white balance.
  • Visual annotations rendered on the photo: grids, crop guides, red strokes/arrows and pinned labels.
  • One-click auto-corrections for exposure, white balance, and suggested crop with adjustable strength.
  • Constrained inpainting to remove small distracting objects while preserving the subject.
  • Before/after preview and small-res export.
  • Short micro-lessons (30–60 words) that explain each detected issue.

Nice-to-have (phase 2)

  • RAW support and higher-fidelity edits.
  • Live guided shooting overlays (rule-of-thirds grid, face placement) in camera preview.
  • Assignment/challenge mode with graded feedback.
  • Side-by-side pro examples and AI-synthesized reworks in a chosen style.
  • Batch analysis / portfolio review.

Pro / advanced

  • Full background replacement with semantic control (respect “no props” constraint).
  • Style transfer / film looks that preserve subject detail.
  • On-device editing option for privacy (smaller models or WASM), with server fallback for high-res results.

How the “red sharpie” visual guidance works (UX)

  • The analysis engine returns problem regions (polygons), suggested actions (crop, exposure +0.7, move left), and severity scores.
  • The renderer overlays semi-transparent red strokes, arrows, circled areas, and short text labels pinned to regions.
  • Hotspots are tappable: tap a red note to reveal the micro-lesson and a preview slider showing that specific correction.
  • Optional animated micro-gifs show a simple “what to do next time” (e.g., step left, lower ISO, change angle).

Example analysis output (JSON you can use in the frontend)

{
  "compositionScore": 62,
  "issues": [
    {
      "type": "rule_of_thirds",
      "severity": "medium",
      "region": [
        [120, 30],
        [320, 30],
        [320, 240],
        [120, 240]
      ],
      "suggestion": "Crop 10% from right and move subject to left third",
      "annotationText": "Move subject to left third"
    },
    {
      "type": "exposure",
      "severity": "high",
      "region": [
        [0, 0],
        [1024, 0],
        [1024, 150]
      ],
      "suggestion": "Reduce highlights by 0.8 stops",
      "annotationText": "Blown highlights"
    }
  ],
  "masks": {
    "subject": "<rle-or-base64-mask>",
    "background": "<rle-or-base64-mask>"
  }
}