Why do some VR videos look different from others or why 6K 60FPS tells you nothing
There are significant differences between videos of the same specs coming from different producers. There’s more to tech set-up of 6K 180º stereo 60FPS that makes videos looks great. The numbers simply tell us that if everything else is done properly the video fits max headset specs.
What happens on the set matters just as much, such as how the filming and lighting is set up, how it’s all wrapped up in post-production and then how the video is brought to the headset, along with our individual eyes and brains that we watch them with. These all have a huge impact on the final video, in parallel to spec numbers.
To look at it in more detail, let’s take these two scenes as reference: Your White Knight from BaDoinkVR and Cheeky from StripzVR, which both seem superior to most other producers. These give a strong sense of presence with actors bodies, objects and colors that accurately reflect real life. The body as an object is distinctive from the background and with a dedicated skin color that looks natural and doesn’t interfere with other colors.
| Reference | Others | |
| Body | Distinctive from the background and with a dedicated skin color that looks natural, doesn’t interfere with other colors, has enough depth to perceive a sense of a real body. Realistic scale | Body is perceived with a little depth, rather as a relief part of the background than standalone object with those background colors reflecting on it. Confusingly looks too big or too small with what you can judge from other details |
| Face | Sharp face details, such as eyes, mouth and nose, look clear | Looks flat, not distinctive, little details |
| Objects | Objects have realistic scale, distinct depth, separation and colors. Sharp or cylindrical objects like arms and legs give perception you can grab these | Objects are overlapping, don’t give perception of standalone objects, shapes are not distinctive |
| Colors corrects in post-production |
Each object has natural distinctive color with clear edge that matches its own shape | Colors are blended creating diffusive edges, interfering with each other |
| Lights (effects of light on the picture) | The scene is properly lighted making everything distinctive and properly colored for the camera. Shadows and reflections on objects contribute to realistic perception of it. Empty space in between walls seems transparent |
The scene doesn’t look distinctive. Empty space in between walls is rather grayish and blurry. |
| Camera angles | Bodies and objects are in expected size and perspective. The scene is seen in a way to let things unfold in a natural way | Body and objects feel distorted and indistinctive. Improper perspective makes it hard for viewer to see all elements of the story |
| Photogrammetry Corrects in post-production |
Straight object like walls or refrigerators have straight shape in VR, convergence point makes it easy for eyes to follow the action. | “Bended” walls, things that are expected to be straight are not so. Eyes get strained fast, it’s hard to look and focus on moving objects. |
| Image | Sharp, good light and colors | Blurred, weak light or overexposed light, unnatural or pale colors |
Take into account that quality often differs a lot from one release to another from the same producers. Newer videos are expected to provide much better quality over these produced in 2016/2017, though with some producers their older videos might look better.
We at SLR attempt to fix the issues in post-production using raw footage from the camera if provided. Another option is to improve things in video playback with the SLR app offsets for each video. Not much could be done with bad setup or something went wrong on the set.
Both the custom camera rig and the manufactured camera used to produce reference videos are also used by other producers with under-performing results. It should be noted that one of these is Z Cam K1 Pro and it always uses the same lens settings and the same software for stitching footage together. This means the difference can only be due to the method of shooting and preparing the lighting for each scene.
Most studios follow this conventional standard: 5K 180º hardware 60FPS recorded with DIY GoPro6/7 stereo rig or 6K 180º hardware 30FPS interpolated to 60FPS 30Mbps (using the ZCam K1 Pro). But as you can see, videos produced on the same hardware by different teams vary a lot in how they look.
We can see that what matters is not just the set-up of the film set and video equipment, or what happens in post-production. Ideally, both of these would be utilized to their very best, but as we can see this is not what usually happens.
These are the factors that matter when it comes to video quality:
- Sensors and lenses (single camera filming hardly matters, but for stereo filming everything needs to be perfectly matched and calibrated)
- Lighting on the set and how objects are positioned (shadows and reflections of light play a key role)
- Rig setup and calibration (it's easy to misalign things when using a moving camera, or even if the rig is touched while filming, and it's difficult to check all the calibrations every time this happens)
- Camera angles and how actors and objects move
- Post-production, color grading, steroscopic alignments and stitching
- Rendering and encoding for each headset. Downloaded and streaming files are also different and there are around thirty settings to consider when preparing for either format
- The headset's hardware decoder, playback engine and the video player settings
- The headset itself (display, lenses)
- Playback method: local file, streaming, viewport streaming with the best quality or WebVR which gives you the lowest possible quality
Each one of these factors has multiple variations and the above list could be easily extended many times over to include everything involved. Even audio affects how you perceive the picture.
For the highest quality SLR offers interpolated to 90FPS 5K videos to be watched on Oculus Rift, HTC Vive and Windows MR headsets to match the 90Hz display rate. 90FPS delivers smoother movement while the picture remains the same. An Nvidia 10xx or 20xx GPU is required for this.
The cameras used in filming have built-in hardware encoders, meaning you never get real RAW files. To get the fisheye image you have to cut out around 2700x2700px or 3000x3000px per eye and align them next to each other. The original bitrate is generally around 120Mbps, which is then compressed to 30Mbps or 50Mbps constant bitrate. At SLR, we believe we get better quality using variable bitrate resulting in a 40% smaller filesize. Bitrate/compression is the easiest part for us to manage - you can read more about this here.
Viewport technology brings the highest technical quality on every headset.
---
Did you enjoy reading this blog post? Do you have a good understanding of how it all works? Apply for a job with SexLikeReal and help bring the world of VR to life!
SLR fans who have joined the company have turned out to be our best employees. We’re a positive, happy company and we all enjoy the work we do.
Mail us at careers@sexlikereal.com with your CV and a few lines about what motivates you when it comes to joining SLR. We are greatly expanding into the VR ecosystem and we’re excited about getting talented people to come and work hand-in-hand with us.
