Contents
Lately I've been digging into a 3D human body model called BodyParts3D.
It started when I saw an AI-built 3D anatomy viewer on X.
How on earth is this made?
Following the data underneath it led me to BodyParts3D, published in Japan in 2009. That's Part 1 and Part 2.
But digging further turned up something else.
This model crossed the sea at some point, and grew enormously.
Its name is Z-Anatomy.
![]()
Z-Anatomy didn't build a human body from scratch
Z-Anatomy is an open-source 3D atlas of human anatomy.
Inside Blender you can examine the structures of the body: bones, muscles, nerves, vessels, an enormous number of anatomical structures.
Seeing it for the first time, you think: did someone model all of this?
They didn't.
What Z-Anatomy stands on is BodyParts3D, made in Japan.
Z-Anatomy's GitHub repository says it plainly. The models are derived from BodyParts3D. The definitions rely heavily on Wikipedia. And at the top of the author list is Kousaku Okubo, credited for the original BodyParts3D model.
So, roughly: Z-Anatomy is BodyParts3D developed into a 3D anatomy atlas that's easier to study and browse.
The story starts in Japan
BodyParts3D is a 3D human body database built by Japan's Database Center for Life Science (DBCLS) and others. Published in 2009.
That year matters to me.
- No generative AI. No ChatGPT. No Claude. Certainly no "hand it an image and it makes a 3D model".
Work that today looks like a few minutes of prompting had to be done by people, patiently: building the data, organising it, checking it.
Turn the enormous number of structures in a human body into 3D data. Name them. Make a database. Publish it. And get it into a state where other people can use it.
Behind data we now download with one click sits an absurd amount of time.
As I wrote in Part 2, on the underlying TARO model the tissue identification alone took more than three years.

And then it crosses the sea
One of the people who found BodyParts3D was Gauthier Kervyn.
His path is a good one. According to his speaker profile for Blender Conference 2022, he came from visual arts — comics — and then moved into orthotics and prosthetics, learning 3D in order to produce 3D-printed orthoses.
After making hundreds of 2D anatomy illustrations for a company, he started building the first open-source atlas of human anatomy in Blender.
What he built it from was BodyParts3D, published in Japan.
Not "rebuild", but "carry on"
This is the part I like.
Rather than modelling an entire human body again from nothing, use the excellent thing that is already public, and make the next value on top of it.
Kervyn organised the body's structures in Blender on top of BodyParts3D, and rearranged them according to Terminologia Anatomica, second edition (TA2, 2019), the international anatomical terminology.
The file published on Zenodo in June 2021 says the same: it's a modified version of the original dataset shared by BodyParts3D, organised according to TA2 within Blender. The authors are listed as Kousaku Okubo and Gauthier Kervyn.
BodyParts3D × Blender × TA2. That combination becomes Z-Anatomy.
Then the engineers arrive
Here's where it gets interesting. The project didn't stay one person's work.
The credits on GitHub divide the roles clearly.
| Role | Person |
|---|---|
| Original model (BodyParts3D) | Kousaku Okubo |
| Design, 3D, anatomy | Gauthier Kervyn |
| Blender add-on (Python script) | Marcin Zielinski |
| Unity development | Lluís Vinent |
| Translation of anatomical structures | Contributors for Portuguese, Spanish, Italian, Polish and Persian |
Translators show up too. Portuguese. Spanish. Italian. Polish. Persian. People around the world joining in, a piece at a time.
The Blender Conference 2022 description puts it like this: in less than two years, Z-Anatomy gathered more than 5,000 three-dimensional anatomical structures and labels with more than 3,500 definitions, in two fully functional open-source viewers. And: this is how a Belgian medical illustrator, a Polish Python coder and a Spanish developer joined their efforts.
3D human body data born in Japan, grown by people all over the world.
Isn't that a great story?
BodyParts3D → Z-Anatomy → and 2026
The whole thing on one page:
The years are approximate, based on the distribution pages and presentation materials. There are more translators and reference models than listed here (see the credits on GitHub).
View as a table
| When | Where | What happened |
|---|---|---|
| 2009 | Japan | DBCLS and others publish a 3D human body database, in a form anyone can use |
| 2010s | Belgium | Gauthier Kervyn, who moved from visual arts into orthotics and learned 3D, finds BodyParts3D |
| 2021 | Blender | Reorganised in Blender according to TA2 (2019) and published on Zenodo, credited to Kousaku Okubo and Gauthier Kervyn |
| 2021–22 | Europe | Marcin Zielinski builds the Blender add-on, Lluís Vinent the Unity application |
| 2022 | Worldwide | Translated into Portuguese, Spanish, Italian, Polish and Persian. In under two years it gathered over 5,000 structures and 3,500 definitions |
| 2026 | Japan | Seeing an AI-built 3D anatomy viewer on X, following the trail leads back to data that started in Japan |
Looked at that way, "a service called Z-Anatomy was built" doesn't begin to cover it. There is a much longer history behind it.
And in 2026, AI is how I find out
Here's the part I enjoy most.
What led me to BodyParts3D was AI.
I saw someone abroad showing an anatomy model in an interesting way using AI, on X. "What is this?" So I looked it up. There's a 3D model underneath. Follow it further. There's Z-Anatomy. Follow it further. Back to BodyParts3D, in Japan.
In 2009, Japanese researchers made something. People abroad found it, modified it, organised it, turned it into software and spread it around the world.
And in 2026, thanks to AI, someone in Japan finds out about it.
What a loop.
And I get to make a little of the next bit
Since I'm here, I've been trying things with the model myself.
I'm not a 3D modeller, and I'm not an anatomy database researcher. Which is why I'm looking at the UI and the learning experience instead.
- Could structures be easier to explore?
- Could it be easier to study on a phone?
- Could it combine with teaching content?
- What happens if you combine it with AI?
That kind of thing.
While working with the data, I found classifications in the copy I'm using where a muscle sits in the bone category, so I'm currently going through the categories and labels as well.
Not to declare the original anatomy model wrong, but to check, organise and adapt it for the environment I'm using it in.

Maybe this is what open source actually is
Looking into this changed how I see open source a little.
It isn't "the free stuff".
Someone spends time making something, and publishes it. Someone else finds it. They add new value. They publish that. Someone else again carries it on.
Work handed from person to person.
If BodyParts3D hadn't been published, Z-Anatomy might be a different thing today. If Z-Anatomy hadn't been published, I might not be able to handle 3D anatomy data this easily.
And what I make might one day feed someone else's idea.
That may be the real fun of publishing openly.
One thing worth checking before you carry something on
This relay only works because of the licence notices.
On Z-Anatomy's GitHub, Z-Anatomy itself is under CC BY-SA 4.0, and the original BodyParts3D model is to be attributed as "CC-BY-SA 2.1 Japan". Meanwhile the file distributed on Zenodo is listed under CC BY 4.0.
The notice differs by where you get it, so check the licence on the page you actually download from. If you received it under a share-alike condition, what you publish carries the same condition.
Not "free, therefore anything goes", but "free, if you keep the conditions". Same root as the copyright post.
Precisely because it's the AI era, I want to trace things back
Since AI arrived, the speed of making things has become absurd. An app in a few hours. Images. 3D. Text.
Which is exactly why I don't want to forget the question: where did this material come from?
The body model that appears on screen in an instant: follow it back and there's someone who built the data in Japan in 2009. Someone abroad who found it. Someone who organised it. Someone who wrote the code. Someone who translated it. And someone using it now to make something new.
AI isn't a thing for erasing that history. It's a thing you can use to make the next part of it. That's how I've come to see it.
Grateful for an era where you can meet the world on X
One more thing I felt this time.
If X didn't exist. If I hadn't seen that post from abroad. If I hadn't got curious about their idea. I don't think I'd have looked into BodyParts3D this far.
Learning about something born in Japan from someone abroad. Looking into the background and coming to appreciate the work of Japanese researchers. And then using it to make something myself.
Different countries. Different languages. Different fields. And still, one idea connects people.
I'm grateful to live in a time when you can meet people all over the world through X, trade ideas and learn from them.
And I intend to make a small piece of the next leg of this long relay myself.
The series
- Part 1: I found a 3D human body viewer built with GPT-6 Astra, so I translated it into Japanese
- Part 2: Who actually made this 3D human body? Following the trail led to Japanese radio-wave research
- Part 3: this post
Sources
- Z-Anatomy on GitHub (credits and licence)
- Z-Anatomy at Blender Conference 2022
- Open source atlas of human anatomy, Zenodo (15 June 2021)
- BodyParts3D / Anatomography
The people, roles and scale described here were checked against the sources above on 8 September 2026. Licence notices differ between distribution points, so always check the licence on the page you download from.