Estimate how likely your baby is to have brown, hazel, green or blue eyes. Add the grandparents’ eye colours and the estimate gets sharper — they are what reveal the colours a parent carries without showing.

Written by Debbie B. Harper · Founder, My Baby Generator
Many babies are born with blue or grey eyes regardless of what they will end up with. The iris gets its colour from melanin, and a newborn has laid down very little of it — the pigment cells are there, but they largely start working after birth, prompted in part by light.
Colour usually settles somewhere between six months and three years old. Most of the change happens in the first year, and by the first birthday a lot of children are close to their adult colour; the drift can still continue quietly until around three. Darker outcomes tend to show earlier, because they need the most pigment and the change is the most visible.
A practical way to read it: at three years old, the colour you see is very likely the one that stays. Before six months, it tells you little. This calculator predicts the settled colour, not the one in the hospital photograph — which is why a newborn with blue eyes does not contradict a result that gave blue a small chance.
Two things are worth a doctor’s appointment rather than a calculator: eyes that are visibly two different colours, and a change that happens suddenly rather than over months. Both are usually harmless, and both are worth having looked at.
Four steps, in order. There is no randomness in any of them, which is why the same colours always give the same percentages.
The first gene decides whether the iris makes a lot of brown pigment. The second decides whether it lays down the yellow-brown pigment in between. Brown eyes have both. Hazel has the first without the second, green the second without the first, and blue has neither.
Everyone holds two copies of each gene and shows only the stronger one, so brown eyes hide either two brown-promoting copies or one of each. Your eye colour narrows this down; it does not settle it.
This is the step other calculators skip. A parent got one copy from each of their own parents, so the grandparents’ colours rule genotypes out: a brown-eyed parent with a blue-eyed mother or father must be carrying a blue-promoting copy, because a blue-eyed parent had none of the other kind to give.
Each parent passes on one copy of each gene, picked at even odds from the two they hold. Every combination the pair can produce is worked through, grouped by the colour it would give, and the totals become the four percentages above.
Eye colour used to be taught as one gene with brown beating blue. That is now known to be wrong. Current genetics attributes eye colour to something in the region of sixteen genes — HERC2, OCA2, SLC24A4, TYR and others — interacting to control how much melanin ends up in the iris and how it is distributed.
This calculator uses two of them, and it uses them because they carry most of the effect: a variant near HERC2 that regulates OCA2 accounts for a large share of the brown-versus-blue difference in people of European descent. Two genes reproduce the broad pattern well. They do not reproduce every family.
Three consequences worth stating plainly, because they change how you should read the percentages:
For the genetics itself, rather than our summary of it, the National Institutes of Health publishes a plain-language account at MedlinePlus Genetics, which is the source this page follows.
This calculator is offered for education and entertainment. It is not a genetic test, it is not a diagnosis, and it is not medical advice. It reads nothing about you, analyses no sample, and knows only the colours you typed into it. Nothing here should be used to make a decision about your health, your pregnancy or your family — for that, speak to a doctor or a genetic counsellor, who can look at your actual history. If you have a specific concern about a child’s eyes or vision, see a clinician rather than a web page.
Nobody can tell you that, and a calculator that names one colour is overstating what genetics allows. What can be estimated is how likely each colour is, which is what the four percentages above are. Where one outcome comes out well ahead, that is the single most likely result — not a settled one.
Because a brown-eyed parent comes in two kinds that look identical. One carries two brown-promoting copies and can only pass that on; the other carries one of each and can pass on either. Their own parents are what separate the two: a blue-eyed grandparent had no brown-promoting copy to give, so a brown-eyed child of theirs must be carrying the other kind. Filling in the grandparents can move the odds of a blue-eyed baby by several times.
Yes, and it is worth entering. Each grandparent is treated separately: one known colour narrows that parent’s possibilities and the other side falls back on population averages. Only the two parents’ colours are required.
Because this model cannot tell grey from blue. Both are irises with little melanin, and what separates them is how the front layer scatters light rather than which genes are involved. Grey can be entered — it is treated exactly as blue would be — but reporting a separate percentage for it would be inventing a number the model cannot produce.
Not yet. Most babies are born with very little pigment in the iris, so blue or grey at birth is common whatever the final colour turns out to be. Colour generally settles between six months and three years, with most of the movement in the first year. The percentages here describe the settled colour, not the newborn one.
Yes, though it is uncommon. This calculator will show 0% for it, because under two genes there is no brown-promoting copy for either parent to pass on. Real eye colour involves around sixteen genes, and that is where the exceptions come from. It is the clearest example of the difference between “this model gives 0%” and “this cannot happen”.
From how common each gene copy is across the US population, which is the only thing available when nothing else is known about a family. It is an assumption, and it is the weakest part of the calculation. Every grandparent you fill in replaces a piece of that assumption with something real about your family, which is why the tool asks.
Because the tool calculates rather than guesses. The same six colours always produce the same four percentages, in the same order. Nothing is random, which is what lets the page show its working.
No. It analyses no sample and reads nothing about you — it works only from the colours you selected. It is a teaching model, offered for education and entertainment, and it is not medical advice. A genetic counsellor or a doctor can tell you things this page cannot.
MyBabyGeneratorA gentle, private glimpse of your future little one — blended from both parents by AI, in under a minute.