What Will My Baby Look Like?
Quick answer: Nobody — and no algorithm — can actually predict your baby's face; it comes down to a genetic lottery running over thousands of genes. But AI can do something almost as fun: blend one photo of each parent into a photorealistic portrait of an imagined child. You can see yours free, right here in the browser, in about a minute.
| Can it be predicted? | No — appearance is polygenic (many genes per trait) |
|---|---|
| What AI does instead | Blends both parents' features into a believable imagined child |
| Features carried over | Eye color & shape, nose, smile, chin, hair, skin tone |
| Time | About 30–90 seconds |
| Cost | First preview free; credits from $2.99 |
It's one of the first questions every expecting couple asks — and one of the few that has no scientific answer. Your baby's face will be built from an enormous combinatorial draw: each of you passes on one of two versions of every gene, and the traits people most want to know about are shaped by hundreds of genes interacting. That's why "baby predictor" tools can't predict anything real. It's also why the honest version of this question — what would our child look like? — is so much fun to explore with AI.
The honest answer: nobody can predict it
The classic school lesson — brown eyes dominant, blue recessive — turns out to be an oversimplification. Eye color is influenced by several genes, with the OCA2/HERC2 region doing much of the work but not all of it. Two blue-eyed parents usually have a blue-eyed child, but exceptions are documented. The same goes for dimples, cleft chins and earlobe attachment: the "dominant or recessive" framing is a useful myth, not a formula. Add the fact that each parent carries two versions of every gene and passes on a random one, and the space of possible children is astronomically large. No photo-analysis tool, and not even a full DNA scan, computes the face.
What AI can do instead — and why it feels personal
Old "baby generators" morphed two faces toward an average, which is why they produced odd, generic results. Current multi-image models — the same class behind the AI family-photo trend people share on TikTok — take a different approach: they generate a brand-new child whose features are inherited from both reference photos. The AI combines your partner's eye shape with your smile, mixes the two skin tones and hair colors, and renders the result in a real scene. It's an imagination, not a forecast — our results are labeled fictional persons — but a believable one, which is exactly why it feels personal instead of generic.
The features parents recognize first
- EyesColor and shape carry over strongly — usually the first thing people check.
- Nose and chinSmall structural cues that make a face read as 'our family'.
- SmileThe smile and expression line are inherited from both photos, not copied from one.
- HairColor and texture blend the two parents; results often surprise people with red or curl.
- Skin toneMixed from both references — the imagined child sits believably between you.
How to see your own imagined baby
Pick one clear, front-facing photo of each parent — good light, no sunglasses. Choose a scene (studio, Christmas, beach, first birthday…) and press generate. In about a minute you'll have a photorealistic portrait of the child you might have, then you can try the baby with Mom or Dad, or the whole family of three. The first preview is free; there's nothing to install.
See what your baby could look like
Two photos in, an imagined child out — first preview free.
Try it free →Reference