# SEE MEE

> SEE MEE is a working Chrome extension demo for visual try-on while shopping online: save one reference photo, browse a store, click SEE MEE on a product image, generate a faithful try-on inline, and keep the result in a private local Closet.

Canonical human page: https://pragmatiko.me/experiments/see-mee

## What It Is

- A Manifest V3 Chrome extension.
- A browser-native shopping tool, not a separate try-on app.
- A demo/prototype with safety boundaries, not a production identity system.
- A practical experiment in private, user-controlled visual personalization.

## User Flow

1. Install the extension.
2. Save one consenting reference photo.
3. Add an image generation API key and model settings.
4. Browse a normal shopping site.
5. Hover over eligible product/model images.
6. Click SEE MEE.
7. Generate one edited catalog-style image.
8. Replace the product image inline on the page.
9. Save successful results locally to a Closet backed by IndexedDB.

## Architecture

- Popup: reference photo, provider/model settings, API key status, quality/size controls, Pre-generate 10, Reset Page, Clear, Copy Status, and Closet.
- Content script: detects product images/cards, injects SEE MEE buttons, handles inline loading/result/error overlays, suppresses retailer hover swaps, and restores saved results.
- Background worker: validates settings, fetches product images, sends image edit requests, returns base64/data URLs, categorizes errors, and records useful debug status.
- Storage: local browser storage and IndexedDB. No backend is required for the demo.

## Model Logic

- The product image is the source of truth for garment shape, fit, color, logos, seams, pose, crop, lighting, and styling.
- The reference photo supplies identity constraints only: face, hair, visible skin tone, and likeness.
- The model should not restyle, redesign, repaint, simplify, crop, or reinterpret the garment.
- One generation at a time. Pre-generate queues sequentially to avoid API spam and broken state.

## Safety Boundary

- Use only your own reference photo or a photo from someone who explicitly consented.
- Do not generate try-ons of people who did not opt in.
- Do not hardcode, log, screenshot, or publish API keys.
- Do not scrape shopping sites in bulk or bypass site restrictions.
- Treat it as a local learning prototype, not a commercial impersonation or surveillance product.

## Why It Matters

SEE MEE tests a bigger idea: AI personalization belongs inside the actual internet surface where decisions happen. The interesting thing is not a novelty face swap. The interesting thing is a private identity layer over commercial pages, controlled by the user.

## Tags

- Chrome extension
- visual try-on
- shopping
- identity
- image generation
- local state
- browser-native AI