Part three was about the tools that actually made the pictures. This part is shorter, and it’s less about code than about a decision I made along the way.

The Kids Player has always been private — your family, your NAS, a password on every new device. So copyright wasn’t really the issue here?

Right, and I should be clear about that up front. Because it’s a private, password-gated app for my own family, I could have reused existing footage — a frame grabbed from a video of the song — the same way I already do for the audio side of the library. Nothing about that would have been a legal problem for a personal app like this.

Then why generate every picture from scratch instead of just grabbing a frame from the video?

Two reasons, and the second one matters more than the first. The practical reason: most of these songs only exist as video, and a single frame rarely shows the whole song. A nursery rhyme has several verses and several little scenes in it — a frame from 0:12 shows one part of the song, not the idea of the song. Picking one frame that represents the whole thing turned out to be harder than it sounds, and a lot of frames I tried just looked like a random freeze-frame, not a picture of the song.

And the second reason?

I wanted the pictures to actually be mine — made the way I imagined the song, with my daughter in mind, not just whatever the original video happened to show. Once I was generating from a description anyway, I could put things in that no stock frame ever would.

Like what?

Two examples I like best. In “The Sun Is Shining” (햇볕은 쨍쨍), there’s a row of kids in the picture — and the one furthest left is drawn to look like my daughter. She knows it, and she loves pointing it out.

In “Cotton Candy” (솜사탕), there’s a picture of a mom holding her child’s hand, walking together. I gave that mom my wife’s face, and the child my daughter’s. When she sees it she says things like “Mom is here” and “I’m here too” — she recognizes both of them immediately.

Was that the plan from the start, or something you landed on along the way?

It came together as I went. Once I realized I could shape every picture myself, putting real faces from our own life into a couple of them was an obvious next step — not something I’d planned before I started.

Is that a one-time thing, or will you keep doing it?

I plan to keep going. These pictures aren’t fixed forever — I expect to revise them again as I find better scenes, and to keep looking at the song list through my daughter’s eyes: what would she like to see herself in next.

Any regrets about how much time the redraws took?

Some of the rounds were tedious, sure — getting a scene, a pose, or a face to look right took real back-and-forth. But a picture that’s actually ours, instead of a random frame from someone else’s video, was worth the extra evenings.

So the pictures exist, some of them are literally her and her mom, and the tab works. What actually happened once she got to use it?

That’s the last part of this series — and it’s the one with the biggest surprise in it.