ZBrush in Historical Reconstruction
This is a making-of article showing a reconstruction made by Pixologic French team member Thomas Roussel. The project was of a Venus head found deep in the Rhone River, in Arles, a city located in France and famous for its old Roman elements. Since the statue's discovery, the Arles museum has performed much research into its original appearance, from the missing parts to the coloration. Additionally, this project has been documented for a TV broadcast aired on a prime time TV show on French national television, showing elements of ZBrush to more than 3.6 million viewers!
Discover how ZBrush was used to work on the scan data of the head, rebuild the figure and finally paint it, in collaboration with the Arles museum and Eclectic Production / 2ASM.
Introduction by Thomas Roussel
I became involved in this project after some interesting discussions at the Japan Expo 2010 tradeshow with François Arnoul, the French distributor of ZCorp 3D printers and scanners. He had to scan an old Venus head requested by the Arles Museum for analyzing, and they were looking for someone to try and reconstruct the missing parts of the head and recreate the painting. The statue was missing nose, lips, chin, eyebrows, parts of the eyes, hair and forehead. Some color pigments had also been discovered, which was surprising for a head which has been under water for such a long time.
This project was exciting because it was far from the "usual" VFX, video games or illustration use of ZBrush and mainly pure sculpting, including a lot of constraints. The model had its existing shape and when reconstructing the missing parts I had to take care to not go in the wrong direction. I couldn't go too far into the wild creatively; it's a reconstruction, after all! But doing this allowed the museum to try out a number of hypotheses without touching the original Venus statue.
On the left, the Venus head found in the Rhone river and on the right, the reference used for the reconstruction. Notice that both heads are very similar, except that the Rhône one is more voluptuous.
3D Scan Cleaning and Beginning the Sculpting Process
Once the 3D scan was imported into ZBrush, I had a mesh of 1.5 million triangles. That's certainly a good amount of polygons, but not something most ZBrush artists would even blink at. The challenge here was that these were simply raw triangles rather than subdivision levels, which means potential slow down. Fortunately, it wasn't an issue at all on my computer and everything ran smoothly from navigation to sculpting. The problem wasn't polygon counts at all...
The real problem was that the triangulated surfaces created by 3D scanners are not very friendly with subdivision surface software like ZBrush when trying to divide the polygons to add more geometry. On top of that there is a stair-stepped layer effect on the scan's surface which had to be removed. The first step was therefore to smooth this effect without losing any important details. There are no secrets to accomplishing this. It was mainly patient use of the Smooth brushes and occasionally the Clay brush to fix small scan errors.
The first evolution of the rough sculpting.
It was very easy to go further than I should have, but on the other side these features were also a good solution to explain some of my choices. I remember perfectly a discussion with Pascale (the museum curator) as she wanted me to refine the nose and have it thinner because what I did was too large in her opinion. I sent her a couple screenshots showing her that I was respecting the marks visible on the statue where was the original nose, and how my sculpt started exactly at these marks. This Venus, as I mentioned earlier is not as thin as our reference and has more fat. It is important to respect this kind of information while sculpting, even if it's easier to say than do!
From a technical standpoint, these last steps were not very hard to do. I was mainly sculpting with the Clay, Move and Standard brushes, combined with masks. I only had to take care to produce a natural look while still respecting the reference.
Adding the Missing Parts
Like I mentioned in the introduction, it was decided to add some missing parts 'around' the existing head in order to provide a better visual effect and feeling of the Venus. This required neck, shoulders and hair braids while respecting the reference Venus. Regarding the braids, the existing hairs and some parts of the neck have marks on them which confirm the hypothesis that braids originally existed at these locations.
As I mentioned above, the strategy to work on a 3D scan is a little bit different from traditional creative sculpting. It's not a creative work and it is crucial to rebuild the Venus as it was in the past. To accomplish this, I chose to through the following steps:
While defining this strategy and doing the first tests, I couldn't see the original head. I had to talk with the museum about the references and validate some steps by phone. I only had access to photos as references, based from another Venus which is in the Louvre museum reserves.
On top of that, I also had to record the entire sculpting process for the TV broadcast, which I was afraid would not be very helpful for my computer's performance! Thanks to ZBrush, I didn't feel any slow down at all.
The 3D scan topology with a close up. Notice the layer effect.
Some artifacts to be cleaned on the model. The most problematic was the one shown by the left arrow, which had weird topology.
Once this step was finished, I begin to build the missing parts. For this I still used the Clay brush and sometimes the Claybuildup brush. In this way I added the nose, lips, chin, eyebrows, parts of the eyes and hair.
Adding to the complexity was that it is typical to work symmetrically when sculpting a digital figure and then add asymmetry at the end. For example, it is a lot easier to sculpt a nose in symmetry and then apply deformation or a pose when done. That was not possible with this project, which was a model that had already been posed two thousand years before I got to work on it and so was asymmetrical from the start. I had to take great care with this specific point, which hounded me through all the reconstruction steps. This was in fact very good motivation. It reminded me of my clay sculpting training: No undos, no 3D layers and no symmetry!
Speaking of layers, I used them for almost all stages, which allowed me to toggle with a single click between the original mesh and my sculpt. It was invaluable for comparing my sculpting to the original to ensure that I stayed true to the source. I also used ZBrush's TimeLine feature, where I stored key frames for each layer at value 0 or 1 which I could then move between. These ZBrush features were very important as I had to regularly check in with the Museum to see if the progress was in a good direction or not.
The evolution of the sculpting, shown with the help of 3D Layers.
To do this part of the project, I added several ZSpheres as SubTools which I put approximately where I needed them to be. I then refined the position of each piece before converting the ZSpheres to fresh polygons through Adaptive skinning. With the hair braids, I tried to put these ZSpheres as close as possible to the reference. For the neck and shoulders, it was more roughed out, since the purpose was only to have some base material to sculpt with. I then began to sculpt and move the new polygons about as needed before going to the next stage.
A. On the left, the ZSpheres and on the right, the polygons created by the Adaptive Skin.
B. The sculpting is roughly close to the final model as the main volumes and parts are now where they have to be, although it will need a lot of small refinements. Looking back at this sculpting now, after some months have passed I would like refine the shoulders which are too small.