Blog Layout
Fovotec collaborates with future-of-work platform Bublr to improve metaverse experience in WebXR
January 17, 2022
Bublr is an innovative virtual meeting space designed to make collaboration and social interaction more natural and intuitive. The platform represents users with live video avatars that can freely explore 3D spaces in real time and connect with others to meet, work and play.
It’s crucial for a successful metaverse experience that the images should look as natural as possible. Unfortunately, the linear perspective geometry used to render most 3D spaces creates unnatural-looking distortions, especially at wide fields of view.
As the team at Bublr realised, this problem is especially severe when it comes to representing human faces. We might not notice or care if a desk looks a bit stretched, but when it’s our own face, or that of a colleague, then we do.
When Bublr reached out to Fovotec we were happy to let them know that there is a solution. The FovoRender process bypasses the normal linear perspective calculations that standard 3D engines use and replaces them with a novel, patent-protected set of calculations based on the structure of human vision.
In practice this means that a version of Bublr running with Fovorender can show much wider fields of view—which improves interaction and navigation—while keeping the natural appearance of objects—including the faces of the avatars themselves.
The FovoRender solution is particularly effective for WebXR delivery as it is designed to work on standard screens—phones, laptops, and large displays—and can be configured to suit the display size.
Bublr and Fovotec are now working together on developing a FovoRender version of the Bublr platform in the
BabylonJS rendering engine and we hope to share more results soon.
Proof of concept images follow showing some of the Bublr 3D assets running within our Unreal 4.27 FovoRender integration at various fields of view from a standard 90 degree horizontal up to an extreme 150 degrees.
In each set of comparisons the faces of the avatars will be more distorted and the center of the 3D space will feel smaller and further away in the linear view compared to the FovoRender version, where both the faces and the 3D space will have a far more natural experience.
Fovotec announce their new 3D rendering technology Dynamic Projection.
‘David Hockney is our English Picasso’. Have you read this article in the Sunday Times?
Karim Zouak is an award–winning writer/director/producer that has been in the business for almost 20 years. Working in both Canada and the United Kingdom, he has honed his craft as writer, director, executive/supervising producer, editor and visual effects artist on commercials, television series, feature films and countless online videos for clients like MTV, Channel 4, Nike, Sony, Philips, Virgin, Blackberry, Unilever and Diageo. His work has been produced and seen internationally in multiple languages by millions of regular viewers, and developing strong creative work remains the heart of his dedication to the craft. Karim has been the first film director to try FovoRender for Unreal 4.27.2 in the context of a virtual production. Karim has been testing FovoRender on his fantastic full CGI film project Recognition, due for release at the end of 2022, teaser image below:
David Baylis is a 3D artist living in Vancouver, Canada (Originally from Lille, France). He works as a Technical Marketing Consultant at Epic Games producing demos, tips & tricks tutorials and webinars. As a car enthusiast David specialises in automotive real-time rendering using Epic’s Unreal Engine. He has been working with Fovotec as a development partner and has recently used FovoRender to create a hyper-realistic shot of an Audi A5 that looks closer than ever to real life. "I have been actively testing FovoRender on my scenes, and it's hard to go back to the vanilla version of Unreal! The biggest difference I noticed was for the interior shots. Extremely helpful to use FovoRender to control the distortion effect of the camera. Amazing plug in!" - David Baylis, Technical Marketing Consultant, Epic Games Whilst Epic’s Unreal Engine 4.27 is designed to produce photorealistic results and contains incredible technology, David finds it can be limited in options for 3D image composition. In standard Unreal you can either zoom in for a tight shot with minimal distortion or zoom out for a wide field of view with lots of distortion, which can be restrictive, particularly when rendering the relatively small interior spaces of cars. Here you can see how FovoRender helped David to fine tune the image composition to give the most realistic result. The biggest visual difference is on the interior shot where he could use FovoRender to tweak individual elements of the images such as the curvature of the windshield, and to show a more realistic width of the wing mirrors. What else can you see that has been corrected?
ThreePointZero creates higher detail ultra-wide angle renders using FovoRender - for high end automotive visualisation. Threepointzero is a software technology company, interacting with companies in the real time computer graphics visualisation industry with offices in London and Brazil. They have recently been working as a development partner with Fovotec on trialling and configuring FovoRender for a high quality automotive project rendering case study. In their line of work, the more details you present on the screen then ultimately the more realistic the result is and that is the challenge they explored with FovoRender. Before partnering with Fovotec, they were using Unreal 4 Normal Cine Camera. In this case, Paul used FovoRender UE4 Path Tracer to fix common distortions created by linear perspective, making the car look even more realistic, both at 120 and 90 degrees FOV. “FovoRender is perfectly suited for rendering wider and higher angle of view stills and it allows us to capture unseen or hard to spot segments of the environment around the car model.” ~Paul Eliasz, Founder The following images show a before / after comparison of the ThreePointZero Mercedes Benz scene rendered in a linear viewport and with an equal field of view using FovoRender.
Fovotec is very excited to announce the start of a new integration of FovoRender into Intel’s OSPRay, a next generation ultra-realistic path tracing based renderer. Intel® OSPRay is an open source, scalable, and portable ray tracing engine for high-performance, high-fidelity visualisation on Intel architecture CPUs. The purpose of OSPRay is to provide an open, powerful, and easy-to-use rendering library that allows one to easily build interactive applications that use ray tracing based rendering (including both surface- and volume-based visualisations). OSPRay (today) is completely CPU-based, and runs on anything from laptops, to workstations, to compute nodes in HPC systems, and in the cloud. In the future, OSPRay will also support GPUs as part of oneAPI cross-architecture programming efforts. It is clear that OSPRay brings some really interesting new possibilities to 3D rendering and following an initial meeting to learn more with Bruce Cherniak, an Intel ray tracing software engineer, at SIGGRAPH 2022. Then, Fovotec set to work looking at the compatibility and potential integration of the FovoRender processes within this exciting new host platform. While there is some way to go to develop a final FovoRender version, below are some of the first proof of concept image compositions created with an initial version of FovoRender for OSPRay. These renders are designed to demonstrate how the FovoRender spatial composition tools can be used by a 3D artist. For example, we show how to create a more natural-looking and spatially accurate render of the interior of a car, in this case a Volvo, than can be achieved with a standard linear render. Images 1 and 2 below show a comparison between a shot created in Intel® OSPRay using a standard linear render and one created using the new trial integration of FovoRender for Intel® OSPRay with some custom settings. Both shots have a horizontal field of view of around 120 degrees, of a model freely available on sketchfab.com .
Fovotec announced today that its new rendering product FovoRender has been selected by leading computer graphics industry magazine Develop3D as one of the top 30 new technologies set to change the world in 2022!
A common request from our industry partners in the visualisation space has been... "can we deliver the awesome flexibility in composition that FovoRender offers in our real-time tools, but packaged up in the highest quality offline renderers instead?" We've listened, and have been working hard on delivering the tools asked for. Fovotec are pleased to announce the first trials of a new evolution of FovoRender that can be used with Maxon’s Cinema 4 D and Octan e to produce path-traced stills and animations. FovoRender is a highly flexible rendering solution designed around human vision that creates much wider fields of view with far less distortion than traditional rendering tools. Until recently, FovoRender had only been developed for Unity and Unreal, both of which are primarily real-time engines but also have some facilities for high-quality offline rendering of the kind used by professional 3D artists and visualisers. The recent move into professional offline environments is a major step for Fovotec, and builds on the work we’ve been doing with studios such as Soluis , CreateCG , and Lightfield , all of whom have used FovoRender to create high-quality stills and animations. FovoRender for Cinema 4D was built using the Open Shading Language and can be set up with a few clicks. The first industry trials of the new process on architectural and automotive visualisation projects are kicking off very soon, watch this space. First example renders are below, and can also be found on our gallery page . Please get in touch for any more info.
One of the biggest challenges 3D artists face is rendering wide fields of view without creating unnatural distortions. This is especially true for architectural visualisers who want to show the details of a design and the wider spatial context that they sit in. The only existing solution is to simulate wide-angle camera lenses, either rectilinear or curvilinear. The shot below shows an archviz scene rendered with an ultra-wide 10mm rectilinear lens, which is equivalent to a field of view of about 135º:
Ben Walker is a seasoned professional in the automotive visualisation industry. Having spent several years as Head of Real-Time Imaging at the Burrows agency working for clients such as Ford, McLaren and VW, he set up his own agency CreateCG in 2020, just as the pandemic was beginning to bite. Fortunately business has been brisk since. CreateCG have benefited, like many other agencies, from the general upswing in demand for virtual content in recent times. A few months ago Ben produced some beautiful architectural renders with the Unreal Engine 4.25 version of FovoRender that we showcased on our blog . And so when we got the new path-tracing version working recently we were keen to see how he might use it for high-end automotive visualisation. Automotive visualisation presents some unique challenges for 3D artists, particularly when it comes to interior shots. The spaces are almost always tight, and it’s very difficult to move the virtual camera without cutting into windows, columns or headrests. As a result, the ‘heights and angles’ options for camera position are limited. So creative teams involved in the production of car interiors tend to rely on a few tried and trusted setups that use narrow fields of view or long lenses. These can be great for showing off the details of instruments and trim but less good for capturing the overall feel of the cockpit. We sent Ben a copy of the FovoRender path-tracer running in Unreal 4.27 and let him experiment for a few days. Then we met up and asked him how he got on. One of the things that Ben noticed immediately about applying FovoRender to make shots of a BMW i3 was the freedom it gave him to render more of the car’s interior at once. “Straight away you get a huge sense of space using FovoRender. You can get a lovely shot with lots in the frame that looks great; it’s a much more natural image” A very common setup in autoviz is the view of the steering wheel and dashboard area from the driver’s side. Getting a generous volume of space in a shot like this without deforming the internal structure is a classic problem. In a typical workflow, the artist might start with a linear perspective lens equivalent to a camera lens of around 18 mm, which produces a shot like this:
Follow Us
Emulating the true experience of natural vision in real-time computer graphics.