HaircutMatch

An AI tool that analyses your face from a photo — detecting face shape, hairline, and hair density — then recommends personalised haircuts and generates AI renders showing how they'd look on you.

Results page — AI renders of 20+ styles across 4 categories, shown on the user's actual face

The problem

Most haircut recommendation content is generic — articles telling everyone with a "round face" to get the same three styles. A useful tool needs to actually analyse the individual: their specific face shape proportions, hairline, density, and what would genuinely suit them.

What I built

The user uploads a photo. The tool does the rest — no quiz, no manual inputs, no "what shape is your face?" dropdown. The full analysis pipeline:

1. Photo is uploaded and sent to the Vision API for face analysis — landmarks, proportions, hairline shape, density estimation
2. Analysis is fed into a structured prompt that produces ranked haircut recommendations with reasoning for each
3. The top recommendations are rendered onto the user's photo using image generation, so they can see approximately how each would look
4. Each recommendation includes a barber instruction guide — what to actually ask for — and paired beard suggestions

The product uses a pay-per-use model: the initial face analysis is free, the AI renders and full recommendation report are a one-off purchase via Stripe. No subscription required.

Technical highlights

Chaining vision analysis → recommendation generation → image generation in a single user-facing flow required careful sequencing and intermediate state management. Each step can fail independently, so the pipeline handles partial results gracefully rather than showing an error if image generation is slow.

Consumer UX at this price point needs to feel instant and magical — long waits kill conversion. Streaming progress indicators and optimistic UI updates keep the experience feeling fast even when the underlying API calls take several seconds.

---

Stack