SkinProof
Skin AI × eCommerce
Is your skincare actually working?
SkinProof measures your skin with YouCam AI Skin Analysis, checks a product's ingredients against your weakest scores before you buy, shows a realistic target with YouCam Skin Simulation, and then tracks the product week by week until it can tell you: keep it, give it time, or stop.
or watch the 2-minute demo video
Checking service…
25 real products ranked against your scan, with ingredient proof, where to buy, and a try-on for tinted ones.
A quick breakout check of four concerns: acne, redness, oil and pores.
Re-scan every 2 weeks. Changes inside the measurement noise don't count.
1 · Scan
Front-facing selfie, hair off the forehead, even daylight, no makeup. Leave a little space around your face.
No selfie handy? Use a sample face (synthetic people, generated with YouCam AI Image Generator):
2 · Shop what fits your skin
The moment before you buy: instead of the trending bottle, see which formulas target what your scan says needs work, and which ones are likely to irritate you.
25 real products with their published ingredient lists, ranked against your measured priorities in code (no AI guesswork). Tap “Why” for the full reasoning.
3 · Check any product
Not in the list? Paste any ingredient list (INCI) straight from a box or shop page. SkinProof matches it against an open, sourced list of 35 skincare actives, uses ingredient order as a rough strength guide, and scores it against your measured priorities.
4 · Is it working? Trials
A trial tracks the concerns a product targets. Re-scan every two weeks in the same light. SkinProof only calls a change real if it is bigger than the measurement noise (shown as a band) and holds for two check-ins. Untargeted control concerns should stay flat, and redness is always watched for irritation.
The two demo trials use synthetic faces: Priya's check-ins were made with YouCam Skin Simulation, Maya's with an image filter that adds redness. Every score shown is a real YouCam Skin Analysis measurement of those images. Your own trials stay on this device.
How it works
YouCam APIs used
- AI Skin Analysis (SD, 15 concerns + skin type, with mask overlays) for every scan and check-in.
- AI Skin Simulation to show a realistic, moderate improvement target for the concerns a product actually targets.
- JS Camera Kit for guided capture (face position, lighting, angle) — fewer bad photos, steadier scores.
- AI Image Generator V2 to create the synthetic sample faces, so the demo uses no real person's face.
- Makeup VTO (skin-smoothing tint) to try a tinted SPF / BB finish on your own photo, or to cover a breakout today.
- Unit System API to protect the demo budget (live scans pause before units run out; the remaining units are shown at the top).
Agent workflow
- Measure (Skin Analysis) → pick your 3 weakest concerns.
- Ground: parse the INCI list and match a sourced knowledge base (no free-text ingredient claims).
- Score fit (benefit vs. cautions for your skin) in code.
- Explain: an open AI model writes the plain-language verdict from those facts only; a code guard discards any draft that names an ingredient the product doesn't contain, and the facts are always shown beside it.
- Visualise the target (Skin Simulation), then track the trial and give a verdict that respects measurement noise.
We measured the noise
Three faces, each re-analysed after +10% brightness with JPEG recompression and after a 3° tilt with a shifted crop (6 re-captures per concern). Largest score change:
SkinProof uses a band of max(3, largest + 1) points. Real life is noisier (other light, distance, phone), so this band is a floor: SkinProof also shows a capture checklist, requires two check-ins in a row, and labels each verdict's confidence.
For retailers: the same fit score can sit on any product page — see the widget on a sample product page.
Not medical advice. SkinProof is a consumer measurement and shopping aid; for pain, cysts or a spreading rash, see a dermatologist.