Making
via Appia
The idea to build a 3D visualization of Piranesi’s Appian Way was born during my two years master at the Albertina Academy in Turin; it was part, among other projects, of a collection of experimental artistic works aimed at investigating the classical Art graphics by modern 3D graphics techniques.
In the years I spent at university I usually liked to explore different techniques, compare them with each other and try to make them dialogue each other in search of new creative stimuli.
This project was meant to recreate a meaningful work, giving it a new light and a personal reading I come up over the years of experience at the Academy.
At the same time it was also a challenge to push my technical skills to the limit. For this reason, the Via Appia is one of the works that more than others best summarize my original path, my ambitions, the amount of passion, effort and patience required when tackling such demanding projects.
Here below, for people having an interest to learn more about the methodology I followed, there are a few notes about the main development phases of the project.
The work was carried out in 2017/2018, using Blender version 2.78 and it has never been modified since then.
For those wishing to go deeper for personal study, I remind you that some working spaces, terminologies and methodologies could be outdated or no longer relevant or practicable, due to the continuous development and renewal of 3D graphics software products.
Camera Match
At first it was necessary to reconstruct the 3D space using the information that can be captured from the etching such as the scale of the objects in the foreground.
Then I compared them with the people and the X and Y axes direction through the surrounding walls which separate the “walkable” area from the one where funeral monuments are located.
Modeling took a few weeks to complete.
The architectural parts, including the various tombs and vases were the first elements to be modeled from the foreground to the backgroud.
Almost all models were created ad hoc by studying Piranesi’s etching in detail.
To keep the working area as fluid as possible and the file optimized I tried to keep the vertices number very low; to do so I used as far as I could Mirror and Bevel modifiers and design simple low poly octagons with a SubdivsionSurface modifier.
Given the large amount of decorations, sometimes quite heavy as in the case of Corinthian columns (300,000 vertices each), I preferred to create groups, connect them to an Empty and finally place them on the scene.
greenery
Once this first phase of modeling was completed, I started to refine the botanical part of the composition.
As you can see, there are cypresses, palms, maritime pines as well as grass and various types of climbing plants; to speed up the operations I used Particle Systems and the SaplingTreeGen add-on, which are very useful when creating realistic and different trees thanks to the parameters at your disposal. The plug-in also enables you to save a preset, which can be randomized several times
sculptures and people
Most statuary and people were modeled through Make Human, an Open Source program that allows with a few steps to obtain human beings with different physiognomy features, a topology in line with their respective UV maps and armor, and excellent compatibility with Blender.
The weakness of the program, however, is the lack of important presets such as hair or beards, which were treated in Blender once the model was imported via Multiresolution and the Sculpt Mode.
Another useful software that sped up modeling was Marvelous, which I adopted to quickly model realistic cloths and clothes getting an excellent yield.
As I mentioned at the beginning, I had never done a work of such proportions on my own before. The scene, once the models preparation was completed, had almost 14 million vertices, and despite the quantity of objects present in the scene (about 850), the Particle systems and the 80 different materials, the file management and the rendering speed were satisfactory.
To further speed up the final calculation, I unchecked the Refractive Caustics parameters and eliminated the Transmission bounces, since I did not have to handle materials such as glass or water on the scene set. The use of an early Blender Denoiser, which was developed and included in the software starting from version 2.78, also proved to be a valid help.
The rendering was processed using Cycles as render engine and the hardware CPU processor, which is why I preferred to use small tiles (64×64 pixels).
Post production was done through simple Photoshop filters.
