Ultrasound scan

Parent 1 background

Hispanic Mexican

Parent 2 background

Hispanic Dominican

Scan Portrait

Drag the slider to compare the scan with the portrait.

Try an example

Demo interface. Example results generated from publicly available ultrasound images.

Same pose. Every family.

Each portrait started as the scan in its corner. The baby keeps the head angle and expression from the scan, while skin tone and features follow the parents' backgrounds.

Newborn portrait generated for parents of Hispanic Mexican and Hispanic Dominican backgrounds — Hispanic Mexican + Hispanic Dominican

The 3D ultrasound scan it was made from — Hispanic Mexican + Hispanic Dominican

Newborn portrait generated for parents of Asian Indian and White British backgrounds — Asian Indian + White British

Newborn portrait generated for parents of Asian Pakistani and Hispanic backgrounds — Asian Pakistani + Hispanic

Newborn portrait generated for parents of Black African and Black Caribbean backgrounds — Black African + Black Caribbean

Newborn portrait generated for parents of White European and White Scottish backgrounds — White European + White Scottish

Newborn portrait generated for parents of Black African and Hispanic backgrounds — Black African + Hispanic

Newborn portrait generated for parents of White American and Asian Chinese backgrounds — White American + Asian Chinese

Newborn portrait generated for parents of Hispanic and Black Caribbean backgrounds — Hispanic + Black Caribbean

How It Works

Upload the Scan

A 3D or 4D ultrasound image, straight from the session.

Choose Backgrounds

Pick each parent's ethnic background from a simple list.

Get the Portrait

A studio-style newborn photo in about 30 seconds, ready to print or share.

Who It's For

Businesses that serve expecting parents

3D/4D Ultrasound Studios

Offer a keepsake portrait alongside every scan session.

Pregnancy and Parenting Apps

A feature expecting parents will want to try and share.

Photo and Keepsake Businesses

Prints, frames, albums and gifts built around the portrait.

Already running in production with our launch partner.

How we built it

Why ordinary AI image tools can't do this

Ask a typical AI image tool to turn a scan into a baby photo and you get a generic, stock-photo baby. The pose is gone, the face doesn't match the scan, and the parents' backgrounds are either ignored or overdone. Ultrasound images make it harder still: they're blurry, tinted, and often have a hand or the umbilical cord in the way.

So we treated it as a research problem, not a prompt to tweak. We designed and tested more than 50 different pipelines, changing the steps and the models in each, and ran every one with different settings. We tried leading frontier models, both open-source and proprietary, and compared every result side by side against the original scan.

The approach that won is a multi-step pipeline, where each step has one job. Together they hold on to the pose and face shape from the scan while producing a realistic photo. Once it gave consistent results across families, we optimized it for production so it runs fast and reliably for real customers.

50+pipelines designed and tested

Open + closedfrontier models compared

Multi-steppipeline, each step with one job

~30sper portrait, start to finish

This is the kind of generative AI work we like: problems where the obvious approach fails. If you have one, tell us about it.

A portrait, not a prediction

Each portrait is an artistic rendering inspired by the scan and the parents' backgrounds. It's a keepsake, not a medical image or a guarantee of how the baby will look. We recommend presenting it to parents the same way.

See Ultrasound to Newborn in Action

Book a demo and we'll walk you through it, then talk honestly about whether it fits your business.

Book a Demo Contact Us