When we first got seriously interested in virtual reality several months ago, we noticed that there was hardly any video content available. Now we know why; because filming in VR is very, very complicated!
The main problem is that there are no off-the-shelve camera rigs available that film on all sides at the same time in stereoscopic 3D. Sure, there are 360 camera rigs for sale, but as I discussed in a previous article, most of these cameras offer appallingly low resolution, significant fish-eye effects, and flat 2D images with no depth in them.
Virtual reality is about making your mind believe you are somewhere else; present in another reality. This requires high quality images which come as close as possible to the real world, and the real world is high resolution and has depth in it.
The only available camera rig capable of filming 360 3D video content seemed to be the 360HEROS GoPro mount. We read very mixed reviews about this rig, and decided that if we were gonna take this project seriously, we would have to build our own camera system. And that’s exactly what we did!
Figuring Things Out
Neither Thierry nor I have any experience creating a camera from scratch, so we embarked on several months of painstaking research to figure out how a 360 3D camera actually works in theory. Because this technique is so new, we had to scour the dark corners of the internet to locate tips and tricks shared on obscure camera forums by pioneers and camera enthusiasts from around the world.
This is where we learned about stereoscopic 3D, interpupillary distance (IPD), parallax, field of view (FOV), video stitching, and many other crucial concepts. Once we felt we had a good understanding of all the core principles, it was time to bring our ideas to life by creating our very first prototype!
Lasers & Wood
By now this was still a self-funded project, so we looked for ways to test our assumptions while keeping costs manageable. That’s why we decided to laser cut our first prototype from wood for under €40, rented 16 GoPros for a day, and used Video Stitch trial software for testing the post-production workflow.
Thanks to our thorough research, this initial prototype worked surprisingly well! We were able to film in 360 3D on the sides, and in 2D on the top and bottom.
The major downside was that the cameras were able to wiggle around in the mount, which made it impossible to create a well calibrated stitch template. The connection to our tripod was simply a very long screw thread, which worked but was not the most stable solution. A bit of wind and the camera swayed all over the place. Not ideal.
Even though we created ample ventilation holes in the mount, the GoPros overheated quite quickly. And while we made sure every USB port on the GoPros were still reachable for charging and file transfer, plugging in 16 USB cables created a gigantic spaghetti of wires. And these were just some of the many rookie mistakes we made.
In the end, prototype numero uno more than delivered in that it allowed us to test and learn, but we needed a more robust solution. We needed a 3D printer, and for that we needed cash. Fortunately we were able to find a €50k investment within a week from two amazing private investors, allowing us to keep the momentum going and progress to the next level quickly.
3D printing sounds like a dream come true for rapid product design, except they forgot to tell us that “rapid” means 31 hours of print time for 1/6th of our camera mount. It’s more fun to look at than watching paint dry, but definitely not any faster. Also, these 3D printers have a million settings and are clearly still a work-in-progress, which is why we managed to break our printer 3 times in just 2 months.
However, despite these setbacks and about 19 designs later, we are thrilled with the end result! This camera fixes all of the problems we had with our wooden prototype, and then some.
Because 3D printing is super accurate the cameras are totally fixed in place now, meaning we can create accurate stitch templates. Heat sinks placed on the camera casings make sure they do not overheat. We also decided to remove the bottom cameras, since all they show is the tripod, and move those cameras to the top of the rig, which in turn allows us to film in 3D on top as well.
All 16 cameras are neatly wired into an industrial grade USB hub, through which we can power the entire rig (mains and car battery) as well as transfer all recordings from the cams to an external hard disk. Our custom Python scripts are then able to handle file management by renaming each video file and placing them in scene folders, which significantly speeds up our workflow.
On top of this, we added an ambisonic microphone to our rig, which allows us to add spatial 3D audio to our productions for an extra layer of realism.
Next on the list of improvements is eliminating the WiFi remote from the equation by switching to a wired, one-button solution instead of having to rely on shaky wireless connectivity. From now on we will also be filming with an externally powered setup instead of using the internal GoPro batteries. This will enable us to film for several hours straight, and further reduces the chance of overheating the cameras.
If you are curious to see what we can deliver with a camera like this, then you might be interested to hear that we just finished our first major cinematic VR production, together with Arlon Luijten and Emansion, and plan to publicly release this content later this month. If you want to be the first to know when this unique 9.30 minutes long piece of VR content hits the app stores, subscribe to our newsletter on the very bottom of this page!