What Photos to Upload (and What to Skip)

About a third of our regeneration requests come down to one root cause: someone uploaded a blurry or awkwardly-lit photo, and the AI didn't have enough to work with. Here's how to avoid that.

What works

Front-facing. The kid looking at the camera, eyes visible, face unobstructed. This gives the model the best read on facial structure.

Good light. Natural light from a window beats almost any indoor lighting. Outdoor shots on overcast days are great — overcast light is soft and even.

Clean background. Not strictly required but helpful. Busy backgrounds can confuse the model about what's the child and what's the environment.

Multiple photos. One is enough. Two or three is better because it gives the model more angles and expressions to triangulate from.

What doesn't work as well

Profile shots. The AI struggles when the face is more than 30 degrees off-center. It can work, but results are inconsistent.

Hats, sunglasses, Halloween makeup. Anything covering the face. The hair should be visible too.

Group photos. Even if you know which kid is yours, the AI may not. Crop to just the child before uploading, or upload a solo photo.

Heavy filters. Beauty filters, stylized Instagram filters, Snapchat dog-ears. They change the face in ways the AI then bakes into the illustrations.

Blurry action shots. A photo where the kid is mid-run, mid-jump, or mid-laugh can be great as a memory but bad as an AI reference. Look for sharpness over personality.

If you don't have good photos

Spend 90 seconds taking a fresh one. Stand them near a window, ask them to look at you, snap three shots. Done. Better than digging through your camera roll for ten minutes.