SEE MEE: See yourself in the clothing from the brands you love as you shop online
Save one reference photo. Shop normally. Click SEE MEE. The product image becomes a generated try-on of you, right where you found it.
Thesis
Try-on belongs inside the shopping flow, not in a separate upload-and-wait box.
Artifact
A Manifest V3 Chrome extension with a content script, background worker, popup setup, injected SEE MEE buttons, image edits, inline DOM replacement, and a local IndexedDB Closet.
Demo video
Screen recordingDemo recording
The real flow: browse a product page, use the injected SEE MEE control, generate a try-on, and keep the result in the shopping context.
Extension screenshots
Supporting detail
Setup
The extension popup stores the reference photo, model settings, API key, and demo controls.
Closet
Generated try-ons save locally inside the extension so the page can restore results later.
What it is
SEE MEE lets you try on clothes from the brands you love while you shop online.
No separate app. No new tab ritual. Just the product page, your reference photo, and one button.
What it does
A content script scans retail pages for product images, filters tiny assets and repeated UI images, then injects a black SEE MEE button onto eligible product cards.
On click, it captures the product image, product URL, title hints, and page context, then hands the job to a background worker so the page can keep behaving like a normal shopping page.
Demo flow
Install extension. Save reference photo. Open Nike, Uniqlo, or Abercrombie. Click SEE MEE. Keep browsing in the same grid.
For demos, Pre-generate 10 queues visible products one by one so latency does not kill the room with a tiny spoon.
Model path
The working path uses OpenAI gpt-image-2 through an image edit flow. The product image is treated as the source of truth for garment shape, colorway, logos, seams, pose, crop, lighting, and styling.
The saved reference photo supplies identity constraints. The prompt asks the model to preserve the clothing and scene while changing only the visible person identity: face, skin tone, hair, proportions, and light matching.
Local state
The popup stores reference status, provider settings, model settings, and generation history locally in the extension. The Closet is not a cloud account pretending to be privacy.
Generated try-ons are indexed locally so the extension can restore results, clear them, or share the current page state without making the user manage files manually.
Where it goes
The current version uses one reference photo and a general image model, so it works but can drift across face, hair, and body identity.
The stronger version uses a reusable identity profile or lightweight per-user adaptation so likeness stays stable across products. The bigger experiment is browser-native personalization: a private identity layer over commercial pages.