Although headus as a company was founded in 1994, we (Jill and Phil) have been working together since 1991.
Jill was designing artwork for Australian Fine China, a ceramic tableware manufacturer, and had started using software to help speed up the process. She was using CDI paintbox software to visualize different color combinations. Prior to using software, each color design would have to be repainted by hand.
The CDI paintbox software only ran on Silicon Graphics workstations, worth several tens of thousands of dollars at the time, so Jill was borrowing time on one of the few SGIs in the state at Curtin University’s Computing Science department (see SGI 4D60/4D70 below).
That’s where Phil was working, and initially he just helped with some 3D modeling and rendering of Jill’s designs using Alias software (see Lammas Render below). As you can see from the image, the artwork is restricted to the rims of the tableware, and this is because the color was being applied using a decal transfer process. The design is silkscreen printed onto a film, which is then applied to the tableware before final firing. In 1992 though, Jill was wanting to apply her designs over a full plate or bowl, and for that the flat decal needs to fold-up to fit the curved shape.
Phil’s solution was to take a 3D mesh of the plate shape, then squash it down and calculate the tension (tearing) and compression (buckling) forces within the mesh that would result. This was indicated with red and blue lines, and the user could then make cuts in the mesh to allow it to push itself out until the tension and compression forces were reduced. The final pattern could then be printed onto the decal, and it would fold-up to fit the plate shape when applied (see Dart Folding below).
Bar Fridge sized SGI
See below for part of a video slideshow we made in 1992 of our work, showing pdeform (for plate deform) in action. And yes, because this was made in the early 90’s, it was on actual video cassette. DVD video wasn’t invented until 1995. As this was recently recovered from a 25 year old tape, its a touch blurry.
Also in 1992, Jill attended the SIGGRAPH conference in Chicago and saw Cyberware demoing their 3D scanners in the exhibition, then visited them in Monterey after the conference. She brought back some scans of herself, along with shells and a cabbage, with the thought that the 3D texture from the scans could be applied as embossing to some of her tableware designs.
This spawned more software development, and a program called ntest (for nurb test) was written to convert the dense 3D scan data into a NURBS surface. That file could be loaded and manipulated in the Alias modeling software we were using back then.
In 1993, we both attended SIGGRAPH to show this software to Cyberware, and they encouraged us to develop it further. It became CySurf, and was bundled with Cyberware scanners thereafter.
In 1994 we founded headus, with the aim to further explore the use of 3D scanners and the software tools required, and in 1995 took delivery of our own Cyberware PS (head) and MS (objects) 3D scanners.
Even though the very first thing we scanned was a teapot, our focus shifted away from tableware design and manufacture and was 100% dedicated to running a 3D scanning service bureau and developing software to make better use of the scan data for ourselves and our customers. We didn’t have a need for flattening plates anymore, so pdeform sat idle in a dusty corner of our source code archive.
Fast forward to 2004, and we had more 3D scanners (a Cyberware WBX full body scanner and a Cyberware M15 hirez object scanner), and had developed CyDir (scan processing) and CySlice (scan surfacing), also sold alongside the Cyberware scanners.
One of the jobs that came through the service bureau mid-way through the year was a character model for Karim Biri called “Leon”. He was heading cgtoolkit, a group developing a muscle plugin for Maya, and Leon was to be used for their demos. The job involved the hirez scanning of the model, followed by SUBD surface fitting and extraction of displacement and normal maps in CySlice.
Normally back then we’d fit the surface, email the mesh to the customer to create the UVs, then extract the texture maps from the dense scan data using their UVs. Somehow though Karim talked us into creating the UVs. We used Maya’s tools, and because we didn’t do this very often and didn’t know any tricks or shortcuts, found it to be a very unpleasant experience, and swore never ever to do UVs EVER again.
CySlice SUBD Surface
Team Bondi Scans
In early 2005 though another job came up that potentially involved lots of UV editing. With our Cyberware WBX, we captured roughly 250 body scans (around 100 costumes with 2 to 3 poses each) of actors in 1940’s period costumers for Team Bondi’s LA Noir game.
Part of the initial testing for that project involved surfacing one suit scan and extracting normal maps. We were worried that we’d have to do this for a couple of hundred scans, which would have been a lot of UV editing. We wondered then if the method we had developed a decade earlier, to apply flat printed decals onto real curved plates and bowls, could be adapted for texture image mapping onto virtual 3D meshes.
And so the first version of UVLayout was developed, not for release, but purely for our own use on the Team Bondi job. In fact we didn’t think anyone else would be that interested in yet another UV editing tool, but we were happy we had something now to do UVs that we knew how to use.
Later in 2005 though we were demoing a Cyberware 3D scanner and our software at GDC (Games Developers Conference), and UVLayout was a small part of the CySlice demo. We still remember the moment when one guy said something to the effect of … “That’s all interesting, but what’s that you just used to make your UVs. That’s very cool”. The feedback we got from those demos made us realize that we weren’t the only people out there who struggled making UVs, and maybe we should release UVLayout as a separate product.
See below for some videos made as the UVLayout prototype was being developed. You should recognize the Leon model from that 2004 scanning job. So we could compare the results with our original attempt, we used him for much of the early testing on UVLayout.