A thousand doors
Updated: Oct 16, 2019
It's been many months since I posted here, having been kept busy with the day job, but also having absorbed the enormity of the supposed sandbox task of pulling together a mixture of real and invented 3D models into something resembling a convincing place.
Crafting a fully realized environment is a job normally undertaken by teams of designers and modelers, sometimes over several years; teams with access to the latest software and high-end graphics workstations. It was naive of me to think that in spare evenings and in weekend gaps between dog walking and cooking dinner, that I might find time to both hone my modelling expertise and simultaneously continue to be in a relationship with my girlfriend! Even big studios now are turning to pre-fabricated assets to flesh out scenery. Why spend many hours creating subtle variations of convincing skyscrapers when you can simply license a kit?
But that is to lose sight (rather early on!) of the original idea, which was merely to organically collect pieces of many jigsaws in order to make a new sort of Soanerian composition. So returning to that theme, whilst on holiday I thought of something more manageable. Something which had a simple sense of purpose and which could accumulate as my leisure time allowed. Thus was born the "Thousand Doors" idea. I love a characterful door as much as the next guy. Okay, probably more than the next guy. Doors are almost infinitely variable but completely ubiquitous. I pass countless doors everyday, many of which catch my eye. Sometime this will be due to the intricate details of the door itself, or other times it will be down to the sense of urban mystery of what might lie behind it. A closed door is a mystery, though perhaps only on a domestic scale.
Building from a technique I originally referred to as "snap, chop, go", it is possible to have reasonably constant production line of models to send to a Sketchfab collection that I've called "Thousand Doors"
After some reflection I've rechristened this workflow to "smash and grab" photogrammetry. Put simply, I take about one and a half minutes of video on the phone, break the video down into two frames per second JPGs (using ffmpeg) and then stick the lot into Agisoft Photoscan. The models aren't archival quality but they are pleasing enough to look at. I generally decimate them down to around 5000 faces - more than enough to make sure the geometry isn't compromised too much, and then I rely instead on good quality 4K textures to get the sense of detail across.
For reasons of copyright, I won't make these models downloadable, but they will hopefully serve as a useful reference resource providing the tagging is consistent.
The wait for the next blog post won't be very long as I have something lined up which owes nothing to my flights of fancy about imagined spaces but is firmly rooted in the practicalities of digital cultural heritage for smaller heritage institutions working on a shoestring budget.