In April of 2017, world famous climber Ueli Steck died while preparing to climb both Mount Everest and Mount Lhotse without the use of bottled oxygen. Ueli’s close friends Jonathan Griffith and Sherpa Tenji attempted to finish this project while director/photographer Griffith captured the entire story. The result is the 3D VR documentary, Everest VR: Journey to the Top of the World. It was produced by Facebook’s Oculus and teased at last year’s Oculus Connect event. Post-production was completed in February and the documentary is being distributed through Oculus’ content channel.
Veteran visual effects artist Matthew DeJohn was added to the team to handle end-to-end post as a producer, visual effects supervisor, and editor. DeJohn’s background includes camera, editing, and visual effects with a lot of experience in both traditional visual effects, 2D to 3D conversion, and 360 virtual reality. Before going freelance, he worked at In3, Digital Domain, Legend3D, and VRTUL.
As an editor, DeJohn was familiar with most of the usual tools, but opted to use Blackmagic’s DaVinci Resolve Studio and Fusion Studio applications as the post-production hub for the Everest VR documentary. Posting stereoscopic, 360-degree content can be quite challenging, so I took the opportunity to speak with DeJohn about using DaVinci Resolve Studio on this project.
[MD] I have had a high comfort level with Premiere Pro and also know Final Cut Pro. Premiere has good VR tools and there’s support for it. In addition to these tools I was using Fusion Studio in my workflow so it was a natural to look at DaVinci Resolve Studio as a way to combine my Fusion Studio work with my editorial work.
I made the switch about a year and half ago and it simplified my workflow dramatically. It integrated a lot of different aspects all under one roof – the editorial page, the color page, the Fusion page, and the speed to work with high-res footage. From an editing perspective, the tools are all there that I was used to in what I would argue is a cleaner interface. Sometimes, software just collects all of these features over time. DaVinci Resolve Studio is early in its editorial development trajectory, but it’s still deep. Yet it doesn’t feel like it has a lot of baggage.
[OP] Stereo and VR projects can often be challenging, because of the large frame sizes. How did DaVinci Resolve Studio help you there?
[MD] Traditionally 360 content uses a 2:1 aspect ratio, so 4K x 2K. If it’s going to be a stereoscopic 360 experience, then you stack a left and right eye image on top of each other. It ends up being 4K x 4K square – two 4K x 2K frames stacked on top of each other. With DaVinci Resolve Studio and the graphics card I have, I can handle a 4K x 4K full online workflow. This project was to be delivered as 8K x 8K. The hardware I had wasn’t quite up to it, so I used an offline/online approach. I created 2K x 2K proxy files and then relinked to the full resolution sources later. I just had to unlink the timeline and then reconnect it to another bin with my 8K media.
You can cut a stereo project just looking at the image for one eye, then conform the other eye, and then combine them. I chose to cut with the stacked format. My editing was done looking at the full 360 unwrapped, but my review was done through a VR headset from the Fusion page. From there I was also able to review the stereoscopic effect on a 3D monitor. 3D monitoring can also be done on the color page, though I didn’t use that feature on this project.
[OP] I know that successful VR is equal parts production and post. And that post goes much more smoothly with a lot of planning before anyone starts. Walk me through the nuts and bolts of the camera systems and how Everest VR was tackled in post.
[MD] Jon Griffith – the director, cameraman, and alpinist – a man of many talents – utilized a number of different systems. He used the Yi Halo, which is a 17-camera circular array. Jon also used the Z CAM V1 and V1 Pro cameras. All were stereoscopic 360 camera systems.
The Yi Halo camera used the Jump cloud stitcher from Google. You upload material to that service and it produces an 8K x 8K final stitch and also a proxy 2K x 2K stitch. I would cut with the 2K x 2K and then conform to the 8K x 8K. That was for the earlier footage. The Jump stitcher is no longer active, so for the more recent footage Jon switched to the Z CAM systems. For those, he would run through Z CAM’s Wonderstitch application, with is auto-stitching software. For the final, we would either clean up any stitching artifacts in Fusion Studio or restitch it in Mistika VR.
Once we had done that, we would use Fusion Studio for any rig removal and fine-tuned adjustments. No matter how good these cameras and stitching software are, they can fail in some situations. For instance, if the subject is too close to the camera or walks between seams. There’s quite a bit of composting/fixing that needs to be done and Fusion Studio was used heavily for that.
[OP] Everest VR consists of three episodes ranging from just under 10 minutes to under 17 minutes. A traditional cinema film, shot conservatively, might have a 10:1 shooting ratio. How does that sort of ratio equate on a virtual reality film like this?
[MD] As far as the percentage of shots captured versus used, we were in the 80-85% range of clips that ended up in the final piece. It’s a pretty high figure, but Jon captured every shot for a reason with many challenging setups – sometimes on the side of an ice waterfall. Obviously there weren’t many retakes. Of course the running time of raw footage would result in a much higher ratio. That’s because we had to let the cameras run for an extended period of time. It takes a while for a climber to make his way up a cliff face!
[OP] Both VR and stereo imagery present challenges in how shots are planned and edited. Not only for story and pacing, but also to keep the audience comfortable without the danger of motion-induced nausea. What was done to address those issues with Everest VR?
[MD] When it comes to framing, bear in mind there really is no frame in VR. Jon has a very good sense of what will work in a VR headset. He constructed shots that make sense for that medium, staging his shots appropriately without any moving camera shots. The action moved around you as the viewer. As such, the story flows and the imagery doesn’t feel slow even though the camera doesn’t move. When they were on a cliffside, he would spend a lot of time rigging the camera system. It would be floated off the side of the cliff enough so that we could paint the rigging out. Then you just see the climber coming up next to you.
The editorial language is definitely different for 360 and stereoscopic 360. Where you might normally have shots that would go for three seconds or so, our shots go for 10 to 20 seconds, so the action on-screen really matters. The cutting pace is slower, but what’s happening within the frame isn’t. During editing, we would plan from cut to cut exactly where we believed the viewer would be looking. We would make sure that as we went to the next shot, the scene would be oriented to where we wanted the viewer to look. It was really about managing the 360 hand-off between shots, so that viewers could follow the story. They didn’t have to whip their head from one side of the frame to the other to follow the action.
In some cases, like an elevation change – where someone is climbing at the top of the view and the next cut is someone climbing below – we would use audio cues. The entire piece was mixed in ambisonic third order, which means you get spatial awareness around and vertically. If the viewer was looking up, an audio cue from below would trigger them to look down at the subject for the next shot. A lot of that orchestration happens in the edit, as well as the mix.
[OP] Please explain what you mean by the orientation of the image.
[MD] The image comes out of the camera system at a fixed point, but based on your edit, you will likely need to change that. For the shots where we needed to adjust the XYZ axis orientation, we would add a Panomap node in the Fusion page within DaVinci Resolve Studio and shift the orientation as needed. That would show up live in the edit page. This way we could change what would become the center of the view.
The biggest 3D issue is to make sure the vertical alignment is done correctly. For the most part these camera systems handled it very well, but there are usually some corrections to be made. One of these corrections is to flatten the 3D effect at the poles of the image. The stereoscopic effect requires that images be horizontally offset. There is no correct way to achieve this at the poles, because we can’t guarantee how the viewer’s head is oriented when they look at the poles. In traditional cinema, the stereo image can affect your cutting, but with our pacing, there was enough time for a viewer to re-converge their view to a different distance comfortably.
[OP] Fusion was used for some of the visual effects, but when do you simply use the integrated Fusion page within DaVinci Resolve Studio versus a standalone version of the Fusion Studio application?
[MD] All of the orientation was handled by me during the edit by using the integrated Fusion page within DaVinci Resolve Studio. Some simple touch-ups, like painting out tripods, were also done in the Fusion page. There are some graphics that show the elevation of Everest or the climbers’ paths. These were all animated in the Fusion page and then they showed up live in the timeline. This way, changes and quick tweaks were easy to do and they updated in real-time.
We used the standalone version of Fusion Studio for some of the more complex stitches and for fixing shots. Fusion Studio is used a lot in the visual effects industry, because of its scriptability, speed, and extensive toolset. Keith Kolod was the compositor/stitcher for those shots. I sent him the files to work on in the standalone version of Fusion Studio. This work was a bit heavier and would take longer to render. He would send those back and I would cut those into the timeline as a finished file.
[OP] Since DaVinci Resolve Studio is an all-in-one tool covering edit, effects, color, and audio, how did you approach audio post and the color grade?
[MD] The Initial audio editing was done in the edit and Fairlight pages of DaVinci Resolve Studio. I cut in all of the temp sounds and music tracks to get the bone structure in place. The Fairlight page allowed me to get in deeper than a normal edit application would. Jon recorded multiple takes for his narration lines. I would stack those on the Fairlight page as audio layers and audition different takes very quickly just by re-arranging the layers. Once I had the take I liked, I left the others there so I could always go back to them. But only the top layer is active.
After that, I made a Pro Tools turnover package for Brendan Hogan and his team at Impossible Acoustic. They did the final mix in Pro Tools, because there are some specific built-in tools for 3D ambisonic audio. They took the bones, added a lot of Foley, and did a much better job of the final mix than I ever could.
I worked on the color correction myself. The way this piece was shot, you only had one opportunity to get up the mountain. At least on the actual Everest climb, there aren’t a lot of takes. I ended up doing color right from the beginning, just to make sure the color matched for all of those different cameras. Each had a different color response and log curve. I wanted to get a base grade from the very beginning just to make sure the snow looked the same from shot to shot. By the time we got to the end, there were very minimal changes to the color. It was mainly to make sure that the grade we had done while looking at Rec. 709 monitoring translated correctly to the headset, because the black levels are a bit different in the headsets.
[OP] In the end, were you 100% satisfied with the results?
[MD] Jon and Oculus held us to a high level in regards to the stitch and the rig removals. As a visual effects guy, there’s always something, if you look really hard! (laughs) Every single shot is a visual effects shot in a show like this. The tripod always has to be painted out. The cameraman always needs to be painted out if they didn’t hide well enough.
The Yi Halo doesn’t actually capture the bottom 40 degrees out of the full 360. You have to make up that bottom part with matte painting to complete the 360. Jon shot reference photos and we used those in some cases. There is a lot of extra material in a 360 shot, so it’s all about doing a really nice clone paint job within Fusion Studio or the Fusion page of DaVinci Resolve Studio to complete the 360.
Overall, as compared with all the other live-action VR experiences I’ve seen, the quality of this piece is among the very best. Jon’s shooting style, his drive for a flawless experience, the tools we used, and the skill of all those involved helped make this project a success.
The article originally written for Creative Planet Network.
©2020 Oliver Peters