This is the build of HairOver from July 29, 2026. The app didn't have its current name yet.
Back then, the header of the style report said Stylora. That was the working name. Weeks later it became HairOver, but the Android package ID on Google Play is still com.proyug.stylora. You can't rename a package ID after it ships, so the old name is still in there.
I'm writing this down because it's the most honest record I have of what the product actually was before the landing page, the marketing and the pricing work. Anyone can say "we built an MVP." This is what ours actually did.
The core flow in one line
Photo → analysis → ranked recommendations → style details → try it on your own face.
The whole first version is built around that line. Every screen in this post belongs to one step of it.
Step 1: The analysis screen
You upload a photo and the app reads it. The loading screen shows the steps it's doing: detecting face shape, checking hair attributes, finding your best matches. It also says it "usually takes about 5 seconds."
Why it looks like this: an AI app has a trust problem in its first ten seconds. The user just handed over a selfie and is waiting. A plain spinner looks like nothing is happening. Showing the steps ("detecting face shape…") tells the user the app is doing something specific with their photo. I think this matters more for a face app than almost any other category, because the user's own face is on screen the whole time.
Step 2: The style report
After analysis you get a Style Report: "Top Picks for You," with a line explaining the filter it used. In the build I'm describing, it said the picks were curated for straight hair and a professional + casual + trendy vibe.
The vibe picker is the last step of a three-step onboarding. You choose Professional, Casual, Trendy or Low Maintenance (more than one is allowed), and the recommendations use that choice.
Why a match percentage: a library of 100+ haircuts is too much to scroll through. A percentage sorts it. It's not a verdict. It puts the styles most likely to suit your face shape and hair near the top.
One thing I notice looking back: the top four cards on that report all said 98% Match. When every card shows the same number, the number isn't helping you choose. I'm writing that down here because it's the kind of detail that only shows up when you look back at your own product with some distance.
Step 3: Style details
Tap a card and you get a full page about that style. There's a description, who it suits (face shapes and hair types), key information, a "celebrity inspiration" note, who it's not ideal for, and how to maintain it.
The "Not ideal for" section is the part I like most. Most hairstyle apps only tell you what looks good. Saying when a cut won't work is more useful and more believable. It's also something you can take to a barber.
Every detail page ends with the same button: Try This Style On Me.
Step 4: Try it on your own face
This is the part people actually want to see. The app generates the chosen style on your photo and saves it to "My Studio."
In the July build, the same person could see themselves with a Cool Taper Fade with Highlights, then with a Curly Pompadour. Same face, two completely different cuts. That comparison is the product. Everything before this step exists to make this step feel personal.
Step 5: Style Studio, library and filters
The first version also had ways to browse and generate without going through the report.
And a home screen that brings it together: your "Style DNA" (face shape and score), recent creations, and picks for you. There's also a separate hair analysis that scores hairline and hairstyle.
What I think this first version got right
- One job per screen. Analysis, report, detail, try-on. Each screen has one action button.
- The user's face is always the subject. No part of the flow asks you to imagine yourself in a model's photo.
- It says what won't work, too. "Not ideal for" makes the app more believable.
These are my reads of the design. They aren't validated by data yet.
What I can't answer yet
That's the gap. The first version shipped with a lot of surface area: analysis, report, details, try-on, studio, library, filters, hair analysis, home. I don't yet have clean numbers on which of these screens people actually use, which ones they skip, and which one makes someone come back.
The question I care about most is: what percentage of people who upload a photo reach their first generated look? That's the activation step, and my marketing plan tracks it as "first generation rate." Until I publish that number, everything in this post is a description of what we built, not proof that it works.
Key takeaways
- The first version was called Stylora; the Play Store package ID still says so.
- The flow is photo → analysis → ranked report → details → try-on.
- Showing the analysis steps and saying which cuts won't suit you are deliberate trust-building choices.
- Identical match scores on every card is a real weakness I spotted looking back.
- A lot got built. What I still need is data on which parts matter.