The reconstructions were placed on the Ostia website with the permission of Marzia Vinci, but they are © 2001 Marzia Vinci.
Marzia Vinci graduated in Classical Topography at La Sapienza, Rome, in 1995 with a work on the IGM 150 II NE, Valmontone: Territorio ed Insediamenti. She moved to Geneva in 1997. She was a collaborator of the Musée d'Art et d'Histoire and the University of Geneva, Classical Archeology Unit, for the exhibition "Ostia, Port de la Rome Antique" held in Geneva from February to July 2001. As member of the scientific committee for the exhibition, she took charge of the pagan cults and of the design and implementation of the interactive kiosks. She developed the Virtual Model and the CGI images of ancient Ostia with 3D tools.
The Virtual Model
The 3D images have been produced using Newtek Lightwave 3D. We used the Modeler for the modeling and Layout 6.5 for the rendering. We recreated about 75%-80% of the total surface of Ostia, that means 181 different objects. Each object represents a defined unit in the project and it could be an Insula (a collection of houses, warehouses, and other buildings), a special object (the Tiber, the sea), a single major monument (like the Capitolium or the Theatre). To speed up the modeling of the geometries we developed a Lightwave script to make programmatically regular buildings and different types of roofs. We also used a commercial library especially for columns and mouldings. The Virtual Ostia folder took about 380 Mb of HD space.
We divided the process into the following steps: Planning, Modeling, Texturing, Scening and Rendering.
During the first step of the entire process we scanned the whole set of maps of Ostia, we created a single huge map of the whole city and we analyzed the complexity. It is during this phase that we decided to develop the two Lightwave scripts to model: the HouseMaker to create basic buildings and the RoofMaker to create roofs. Most of the pictures for the textures have been taken and processed during this phase.
The modeling has been the longest phase of the entire process. We applied the map of Ostia to the XZ window for reference. We set up the parameters of the background picture so the distances are expressed in scale 1:1 (that is, a distance of one meter in the picture is a distance of one meter in Modeler). We developed all buildings starting from the Porta Romana going to the Porta Marina, modeling the insulae that belong to the right side (between the decumanus and the Tiber) and along the decumanus. All buildings have been provided with a regular roof with tiles, in order to make them more realistic. All sharp edges (the edges of the walls, the edges of the windows, etc.) have been smoothed with a bevel operation, so that the buildings appear more realistic. During this phase we modeled all major monuments, paying particular attention to the level of detail necessary for the close rendering (moulding, column details).
We created several realistic textures using pictures we took directly in Ostia. We have taken pictures of Marble, Bricks, Opus Spicatum, Opus Reticulatum, Stones and Street Stones. From these real pictures we produced several seamless and tileable textures, and then we prepared a set of "basic" surfaces for the most common surfaces in Ostia. These surfaces are: windows, wall surfaces with and without brick, glass, stones, grass, etc. Each surface contains at least three layers, the color, the texture and the diffusion map. During this phase we added an additional level of detail to the most important monuments, like the horizontal lines on the columns of the Capitolium and the grid of the floor of the Forum, created with texture maps.
Fortunately, the scening required a reasonable amount of work, because the project foresaw only still images or small videoclips, where it is only the camera that has animated channels. One of the most common problems during the scening was the huge amount of points to be loaded at the same time. For our graphic workstation, a scene loaded in Layout can have about 1,000,000/1,500,000 points maximum. Beyond that limit, Layout became too slow and the rendering of a picture required too much swapping to the HD. For this reason, almost all rendering of the panorama's has been created using the compositing techniques. We produced two or more renderings from the same point of view, but loading different sets of objects. Then we composed the images to produce the final render of the panorama. Another problem of the scening was the position of the lights. We used a point light for the Sun, and we used eight point lights positioned on an octagon around the city to simulate the radiosity effect. At last, we used a distant light pointing upward at 90 degrees to simulate the bounced light from the floor pointing upwards to lit the eventual ceiling. During the phase of the scening we put some characters around the buildings. These characters give to the viewers an idea of the proportion of the buildings.
Rendering / Post Production
During the rendering phase we produced all images of the project. Sometimes, to have the best result, we had to render more than once the same scene using different light parameters. At the end of each render, we retouched the image with Photoshop for color/contrast balancing.