Flaris

The problem

Makeup try-on already exists — but nearly all of it is built by brands, and it exists to sell you that brand's products. Point it at your face and the answer is always a shade you have to buy.

Meanwhile the actual difficulty is technique. A tutorial shows the look on someone else's face, at someone else's skill level, and leaves you to work out the rest.

What I built

• AR try-on per procedure. Not one finished look stamped onto your face, but each individual step, so you can see what a technique does before attempting it.
• Over 700 guided looks and procedures, with step-by-step instructions rather than a filtered result.
• Brand-agnostic. It works with whatever you already own, which is the whole reason it can be honest about technique.
• Personalised to skin tone, face shape and style, with favourites for the looks you return to.

The part that was actually hard

Real-time face tracking has to hold a steady frame rate on mid-range Android hardware, which is what most of the audience actually owns. Fast on a flagship is not a result.

Camera permission on a beauty app invites suspicion, fairly. The data-safety declaration and privacy policy had to be unambiguous about what the camera feed is and is not used for.

Where it is now

Live on Google Play. The iOS build is in progress. The site carries a blog and support material.