Baby Generator AIHow is the mom/dad resemblance percentage calculated?
It's a descriptive estimate of how much visual influence each parent's photo had on that specific generated image — not a genetic calculation, and not based on any DNA or family data. Generate again with the same two photos and you can get a different percentage, because it describes that one result, not a fixed underlying number.
What the number actually is
When you upload a photo of each parent, the app generates an image and separately estimates how visually close that specific result is to each of the two source photos, then expresses that as a mom/dad split (for example, 62% / 38%). This is a property of the image the AI just made, evaluated after the fact, not an input the AI was told to hit or a measurement of anything about the real parents beyond their two uploaded photos.
What it deliberately is not
It is not a genetic resemblance score, it does not use or infer any DNA information (photos don't contain genetic sequence data), and it has no relationship to real inheritance patterns like the ones described on our genetics explainer page. A real baby's resemblance to each parent is not reducible to a single tidy percentage — this number is a fun, descriptive readout of one AI image, not a claim about biology. See how accurate baby generator apps really are for the broader version of this same point.
Why it can change if you generate again
Each generation is a new run of the image model on the same two source photos, and AI image generation is not perfectly deterministic — small variations between runs are normal and expected, the same way asking an artist to redraw something twice produces two similar but not identical results. Because the resemblance percentage is calculated from the specific image that comes out, a new image can come with a different percentage, even from the exact same two uploaded photos. Neither number is "more correct" than the other; they're each describing their own result.
Why we show it anyway
It's one of the most-loved parts of the result for the same reason "who does the baby look like" is a genuinely fun question between two people — a percentage gives that debate something concrete and shareable to look at, even though everyone involved understands it's a playful readout of an AI image, not a lab result. See how the whole flow works for where this fits into a typical generation.
See your own percentage
Free first preview, made from your two actual photos.
Questions
How is the mom/dad resemblance percentage calculated?
The percentage reflects how much visual influence each parent's uploaded photo had on the specific generated image you're looking at, as estimated by the AI model for that generation. It is not a genetic calculation, is not based on any DNA or family-tree data, and can come out differently if you generate again with the same two photos.
Is the mom/dad percentage scientifically real?
No. It's a descriptive, per-generation estimate of visual resemblance in that one AI-generated image, not a measurement of actual genetic inheritance. A real baby's genetic resemblance to each parent can't be reduced to a single percentage from a photo.
Why did I get a different percentage when I generated again with the same photos?
Each generation produces a new AI image, and the model's visual blend of the two source photos can vary between attempts, which changes the estimated resemblance percentage along with it. This is expected — the percentage describes that one specific result, not a fixed underlying number.