The widget is the only part of Mirraya a shopper ever touches. An earlier exploration settled the shell (a focused sheet over the product page); these are three flows inside it. Every mock is live — click the trigger on each store, run the whole flow on desktop and phone, then tell us which felt best.
First, the research the designs are built on: the real conversion barriers (the photo ask, the wait, the trust moment) and the design move each one forces.
Shoppers weigh a vivid risk (my body photo) against an abstract benefit at the moment of the ask. → Show the try-on payoff before asking, and put "never stored" in plain words on the drop zone itself — not in a footer.
A plain privacy sentence measurably increases disclosure; third-party trust seals mostly don't, and over-signaling raises concern salience. → One calm sentence at the point of the ask. No lock icons everywhere.
Multi-step flows convert ~3× single-page (13.9% vs 4.5%); each small step is a micro-yes that makes quitting feel like wasted effort. → Height and weight come before the photo. The current widget leads with its scariest ask.
People value a result more when they watch the effort — even preferring a transparent wait over an instant answer. → Narrate real steps ("measuring 40 landmark points…"), never a bare spinner.
Low body esteem lowers try-on adoption; self-congruent imagery builds trust. → Non-evaluative language ("measure the fit", never "analyze your body"), no model imagery near the ask — and the render preserving your face, pose and background is itself the trust story.
Unclear instructions and after-the-fact rejections are the killers (30% of people retake a selfie 3+ times). → Guided capture: the three rules as a visible checklist before the camera opens, and rejections that say exactly how to fix.
Primed, benefit-framed permission requests roughly double acceptance (~35% → ~70%). → Our own "one photo does both" screen says yes first; the native prompt is a formality.
Field Test 1's misses were capture artifacts — loose tops pulled taut, rolled phones — not model failures. → The widget does pipeline QA before upload: fitted-clothes and framing guidance is conversion work and accuracy work.
B's sheet, split in two: inputs on the left, a live canvas on the right that never leaves — it holds the garment when you open, your photo when you add it, the narrated measuring pass, and finally the render with the size stamped on. The VTON isn't an epilogue; it's the stage the whole flow performs on. On the phone the canvas rides on top of a bottom sheet.
Desktop
Phone
The payoff is visible from second zero (privacy-calculus benefit), and the render lands in a frame the shopper has been watching the whole time — no surprise reveal to miss.
Widest sheet of the three; on cramped desktop PDPs it's a big overlay. The canvas must never look broken when a store has no garment image.
Benefit-before-ask · point-of-ask privacy line · easy inputs first · labor-illusion narration on the canvas.
The inverse reveal order: the sheet is one full-bleed canvas — a mirror. Controls live in a slim bar under it. Your photo fills the mirror, the measuring pass narrates over it, then the render fades in and the size stamps onto the image. The size is the caption; seeing yourself in the garment is the product. This is the "undersells the VTON" instinct taken to its limit.
Desktop
Phone
The strongest emotional arc and the most shareable moment — the stamped render is a screenshot waiting to happen. Nothing on screen competes with the payoff.
The size number arrives last, and render latency (~10s live) sits in the hero position — a slow or failed render hurts more here than anywhere. Needs the size shown the moment it's known, before the render lands.
Benefit IS the interface · labor illusion at maximum · body-esteem care (it's you, unedited, in your own setting).
Full multi-step psychology: height (easy yes) → weight (optional, skippable) → then the photo, introduced by a priming screen with the three capture rules as a visible checklist — so by the time the big ask arrives, the shopper has already said yes twice and knows exactly what a good photo looks like. Feels native on the phone; the desktop sheet runs the same screens.
Desktop
Phone
Directly implements the strongest quantified findings (multi-step ~3× single-page; primed permission ~2× acceptance). Every screen has exactly one job, so mobile ergonomics are perfect.
Most taps to the result, and the try-on payoff is least visible up front — the priming screen has to carry the "why bother" on its own. Impatient repeat shoppers need the skip-ahead (returning state).
Foot-in-the-door sequencing · pre-permission priming · guided capture checklist · progress ownership.