What you do not see on the surface.

Front Page Kid looks simple on purpose. This page shows the rules, the work and the tests behind it.

Open all sections

What it is

A website, not an app.

Front Page Kid runs in any browser, on a phone or a computer. You do not install anything. The sample does all of its work inside your browser. Your photo and your facts are not sent or saved.

The site is a static website (pages made ahead of time, then served as files). Static pages load fast and have very little that can break.

The rules

Six rules we build by.

Six rules for every page.
  1. No invented words. A kid's quote comes only from the parent. We removed sample quotes that we wrote, even for a fictional player.
  2. No invented facts. We removed a statistic that we could not confirm. Claims we could not measure, such as "about one minute", came off the site.
  3. Other children are removed, never blurred. A blur still shows a child who did not agree to be in the picture.
  4. Honest labels. Samples say "sample". Plans say "plan". The About page says what works today and what does not.
  5. Plain English. Visitor text follows the rules of Simplified Technical English (ASD-STE100): short sentences and one meaning for each word.
  6. Everyone can use it. Body text is 17 px or larger, text colors meet the WCAG AA contrast level, controls are built for keyboard use, and motion stops when your device asks for reduced motion.
The hidden work

Small things that took the most care.

Seven details most people will not notice.
  1. Images made in your browser. The card and the front page you download are drawn in your browser at 1080 × 1350 pixels, the size Instagram and Facebook show in full. On a phone, the share sheet opens. Nothing goes to a server.
  2. The pro card look. On the card, the player is kept exactly as shot. Only the field behind him is softened and warmed, like a portrait lens. The front page keeps the real photo, because a newspaper photo must stay true.
  3. The press. The waiting screen is a small printing press drawn in the site's ink-and-paper style. With reduced motion, it shows the same steps without movement.
  4. The slider. Drag it to compare the camera roll photo and the card. It also works with the arrow keys.
  5. No long waits. The film and the opening demo each run in about 6 seconds. The film is a real 6-second cut. Speeding up the video in the browser made Safari drop frames, so we cut a new file instead.
  6. Phone first. Every page was measured at 360 and 390 pixels wide: no sideways scroll, and small tap targets made larger.
  7. Waitlist and feedback. Both save to a spreadsheet the owner controls. A hidden field stops simple spam bots. The script also stops spreadsheet formulas from being typed in through the form, and it skips duplicate emails.
The engine

A first version, tested on real photos.

Real photo tests, with the failures kept in.

The engine is the part that turns a parent's own photo into a card. Today it runs on the builder's own computer, not on this website. It uses two AI models that run locally, so test photos never left that computer:

  1. A person finder (a segmentation model, which outlines each person in a photo).
  2. A fill model (an inpainting model, which paints in the background where a person was).

We tested it on 21 real photos in two rounds. We kept the failures in the results, because they show what to fix next.

RoundMethodUsable results
1Remove people from the full photoAbout 7 of 19
2Crop to the player first, then remove people, their shadows and what they heldAbout 10 of 18

What we learned: finding people is reliable. Filling large gaps is the hard part. Cropping to the card first means most gaps fall outside the card. The open problem is a photo where another person touches the player. The plan for that case is a stronger fill model, or a request for a different photo.

Speed: a finished card took 1.2 seconds for a photo with one player, and 4.7 seconds for a photo where two other people were removed.

How it was built

Every change written down.

From nothing to live in one day, with every change written down.

Built by one parent working with AI assistants.

Time (Arizona)Milestone
October 17, 2026, 9:00 a.m.Started from nothing
October 17, afternoonFirst version live at frontpagekid.com
October 17 to 18, overnightDesign passes, engine tests, waitlist, About page and this page
  1. One shared brief holds the approved positioning, copy and facts. A validator checks it before every build.
  2. A log records every decision, with the date and who approved it. It has more than 45 entries.
  3. The project history has more than 65 saved versions, so any change can be undone.
  4. Each release this weekend passed 11 automated checks and a full production build. The owner approved each release before it went live.
Roadmap

What comes next.

Six steps from sample to real service.
  1. Connect the engine to this website, so a parent can upload a real photo.
  2. Solve photos where another person touches the player.
  3. More card styles: a studio look with team colors, and an illustrated look. We will not use AI to redraw a child's face.
  4. A team page that collects every family's front page. It is a plan today.
  5. Spanish, with a review by a native speaker.
  6. Before a real launch: a child-privacy review (COPPA), automatic deletion of original photos, a delete button for parents and a trademark search.
Vision

How it pays for itself.

Free sample, paid seasons, no ads.
  1. Free sample. Anyone can make a sample front page now.
  2. Paid seasons. Families pay per season. Coaches, leagues and grandparents get their own plans. See Pricing.
  3. No ads. These are photos of children. Ads next to them would break the trust the product depends on.
  4. Open source is being considered. We are considering publishing the engine code. The paid service would be the hosted product. This is not decided.

Make a sample front page · About