Inside view : 3D Strasbourg MMXX's visuals

1. Introduction

MMXX is a music video depicting Strasbourg / France during the COVID-19 lockdown. We tried to portrait a digitalised and distorted city where all human interactions were happening through the help of technology.

In order to get to make this happen I had 2 datasets and a lot of time during the lockdown.

Photogrammetry 3d models : Months before the lockdown I started taking thousands of pictures of the city with the idea of recreating some placed in 3D while still keeping the glitch style of raw photogrammetric processing.
Open data : Our city has a high quality dataset of 3d information available under public licence. [Photomaillage 3D Strasbourg 2018]

2. Lore & creative freedom

Lore-wise, the story is told as viewing an archive in a distant future, some digital reminiscence of a far gone era. The archive itself has been corrupted and is thus a mashup between datasets, misplaced buildings, props and objets and time dislocated elements.
This allowed for some reality-bending interpretation of the city and for quite some creative experiments.

3. Hardware & software

On a more technical level, the rendering was mostly done with the combo cinema4D + [octane render] but some elements were also rendered with the open source 3d modelling software [Blender]. Photogrammetry was done using mostly [Meshlab] and [Capturing Reality].

4. Breakdowns

To get some insights about the process, I’ll go through 3 differents scenes and try to break them down. It’s not going to get too technical but if you want more info feel free to send us a message.

4.1 Intro Scene : example of open data usage

The video starts with a linears fly over the city and relies on the open dataset from our local city of Strasbourg.
The challenge for this part is mainly about the weight of the data and the strain put on our computing resources. The dataset is divided into city blocks, each block is approximately 300K polygons. The horizontal slice is composed of 30 blocks, that’s close to 10 millions polygons which is in itself not that much a burden for a modern computer but add in high quality textures and it all begins to get complicated to work with.

This required some polygon decimation or reduction (depending on the software). As the camera is set quite far away for the models, details are not so important and as long as vertical lines remain vertical the textures give enough credibility the the scene.

The process simply is

import obj file into C4D
convert material to octane format
use polygon reduction set at about 90%
rince and repeat for the next block

4.2 Glitchy Photogrammetry

In order to blur the boundaries between physical objets I merged several objects from very different settings. The following pictures illustrate going from raw photogrammetry data (~ 200 pictures for this particular object ) and then calculating mesh data and textures.
As I’m going for a glitchy look and feel there was no use in cleaning the data and I could just be lazy and let the process go.

I’ve done this process multiple time for different objets and then the fun can begin.
For instance, here are two photogrammetry scans, one from the rhin port the other one of a tagged electric box from the city center.

4.3. Glitching architecture

Let’s see another process. Here the principe is quite simple. We get some open data mesh as we’ve seen before, the idea is to select a line of vertices of the mesh and keyframe their movement, breaking any structure into parts.