Category Archives: VR

Editing 360 Video in VR (Part 2)

By Mike McCarthy

In the last article I wrote on this topic, I looked at the options for shooting 360-degree video footage, and what it takes to get footage recorded on a Gear 360 ready to review and edit on a VR-enabled system. The remaining steps in the workflow will be similar regardless of which camera you are using.

Previewing your work is important so, if you have a VR headset you will want to make sure it is installed and functioning with your editing software. I will be basing this article on using an Oculus Rift to view my work in Adobe Premiere Pro 11.1.2 on a Thinkpad P71 with an Nvidia Quadro P5000 GPU. Premiere requires an extra set of plugins to interface to the Rift headset. Adobe acquired Mettle’s Skybox VR Player plugin back in June, and has made it available to Creative Cloud users upon request, which you can do here.

Skybox VR player

Skybox can project the Adobe UI to the Rift, as well as the output, so you could leave the headset on when making adjustments, but I have not found that to be as useful as I had hoped. Another option is to use the GoPro VR Player plugin to send the Adobe Transmit output to the Rift, which can be downloaded for free here (use the 3.0 version or above). I found this to have slightly better playback performance, but fewer options (no UI projection, for example). Adobe is expected to integrate much of this functionality into the next release of Premiere, which should remove the need for most of the current plugins and increase the overall functionality.

Once our VR editing system is ready to go, we need to look at the footage we have. In the case of the Gear 360, the dual spherical image file recorded by the camera is not directly usable in most applications and needs to be processed to generate a single equirectangular projection, stitching the images from both cameras into a single continuous view.

There are a number of ways to do this. One option is to use the application Samsung packages with the camera: Action Director 360. You can download the original version here, but will need the activation code that came with the camera in order to use it. Upon import, the software automatically processes the original stills and video into equirectangular 2:1 H.264 files. Instead of exporting from that application, I pull the temp files that it generates on media import, and use them in Premiere. (C:\Users\[Username]\Documents\CyberLink\ActionDirector\1.0\360) is where they should be located by default. While this is the simplest solution for PC users, it introduces an extra transcoding step to H.264 (after the initial H.265 recording), and I frequently encountered an issue where there was a black hexagon in the middle of the stitched image.

Action Director

Activating Automatic Angle Compensation in the Preferences->Editing panel gets around this bug, while trying to stabilize your footage to some degree. I later discovered that Samsung had released a separate Version 2 of Action Director available for Windows or Mac, which solves this issue. But I couldn’t get the stitched files to work directly in the Adobe apps, so I had to export them, which was yet another layer of video compression. You will need a Samsung activation code that came with the Gear 360 to use any of the versions, and both versions took twice as long to stitch a clip as its run time on my P71 laptop.

An option that gives you more control over the stitching process is to do it in After Effects. Adobe’s recent acquisition of Mettle’s SkyBox VR toolset makes this much easier, but it is still a process. Currently you have to manually request and install your copy of the plugins as a Creative Cloud subscriber. There are three separate installers, and while this stitching process only requires Skybox Suite AE, I would install both the AE and Premiere Pro versions for use in later steps, as well as the Skybox VR player if you have an HMD to preview with. Once you have them installed, you can use the Skybox Converter effect in After Effects to convert from the Gear 360’s fisheye files to the equirectangular assets that Premiere requires for editing VR.

Unfortunately, Samsung’s format is not one of the default conversions supported by the effect, so it requires a little more creativity. The two sensor images have to be cropped into separate comps and with plugin applied to each of them. Setting the Input to fisheye and the output to equirectangular for each image will give the desired distortion. A feathered mask applied to the circle to adjust the seam, and the overlap can be adjusted with the FOV and re-orient camera values.

Since this can be challenging to setup, I have posted an AE template that is already configured for footage from the Gear 360. The included directions should be easy to follow, and the projection, overlap and stitch can be further tweaked by adjusting the position, rotation and mask settings in the sub-comps, and the re-orientation values in the Skybox Converter effects. Hopefully, once you find the correct adjustments for your individual camera, they should remain the same for all of your footage, unless you want to mask around an object crossing the stitch boundary. More info on those types of fixes can be found here. It took me five minutes to export 60 seconds of 360 video using this approach, and there is no stabilization or other automatic image analysis.

Video Stitch Studio

Orah makes Video-Stitch Studio, which is a similar product but with a slightly different feature set and approach. One limitation I couldn’t find a way around is that the program expects the various fisheye source images to be in separate files, and unlike AVP I couldn’t get the source cropping tool to work without rendering the dual fisheye images into separate square video source files. There should be a way to avoid that step, but I couldn’t find one. (You can use the crop effect to remove 1920 pixels on one side or the other to make the conversions in Media Encoder relatively quickly.) Splitting the source file and rendering separate fisheye spheres adds a workflow step and render time, and my one-minute clip took 11 minutes to export. This is a slower option, which might be significant if you have hours of footage to process instead of minutes.

Clearly, there are a variety of ways to get your raw footage stitched for editing. The results vary greatly between the different programs, so I made video to compare the different stitching options on the same source clip. My first attempt was with a locked-off shot in the park, but that shot was too simple to see the differences, and it didn’t allow for comparison of the stabilization options available in some of the programs. I shot some footage from a moving vehicle to see how well the motion and shake would be handled by the various programs. The result is now available on YouTube, fading between each of the five labeled options over the course of the minute long clip. I would categorize this as testing how well the various applications can handle non-ideal source footage, which happens a lot in the real world.

I didn’t feel that any of the stitching options were perfect solutions, so hopefully we will see further developments in that regard in the future. You may want to explore them yourself to determine which one best meets your needs. Once your footage is correctly mapped to equirectangular projection, ideally in a 2:1 aspect ratio, and the projects are rendered and exported (I recommend Cineform or DNxHR), you are ready to edit your processed footage.

Launch Premiere Pro and import your footage as you normally would. If you are using the Skybox Player plugin, turn on Adobe Transmit with the HMD selected as the only dedicated output (in the Skybox VR configuration window, I recommend setting the hot corner to top left, to avoid accidentally hitting the start menu, desktop hide or application close buttons during preview). In the playback monitor, you may want to right click the wrench icon and select Enable VR to preview a pan-able perspective of the video, instead of the entire distorted equirectangular source frame. You can cut, trim and stack your footage as usual, and apply color corrections and other non-geometry-based effects.

In version 11.1.2 of Premiere, there is basically one VR effect (VR Projection), which allows you to rotate the video sphere along all three axis. If you have the Skybox Suite for Premiere installed, you will have some extra VR effects. The Skybox Rotate Sphere effect is basically the same. You can add titles and graphics and use the Skybox Project 2D effect to project them into the sphere where you want. Skybox also includes other effects for blurring and sharpening the spherical video, as well as denoise and glow. If you have Kolor AVP installed that adds two new effects as well. GoPro VR Horizon is similar to the other sphere rotation ones, but allows you to drag the image around in the monitor window to rotate it, instead of manually adjusting the axis values, so it is faster and more intuitive. The GoPro VR Reframe effect is applied to equirectangular footage, to extract a flat perspective from within it. The field of view can be adjusted and rotated around all three axis.

Most of the effects are pretty easy to figure out, but Skybox Project 2D may require some experimentation to get the desired results. Avoid placing objects near the edges of the 2D frame that you apply it to, to keep them facing toward the viewer. The rotate projection values control where the object is placed relative to the viewer. The rotate source values rotate the object at the location it is projected to. Personally, I think they should be placed in the reverse order in the effects panel.

Encoding the final output is not difficult, just send it to Adobe Media Encoder using either H.264 or H.265 formats. Make sure the “Video is VR” box is checked at the bottom of the Video Settings pane, and in this case that the frame layout is set to monoscopic. There are presets for some of the common framesizes, but I would recommend lowering the bitrates, at least if you are using Gear 360 footage. Also, if you have ambisonic audio set channels to 4.0 in the audio pane.

Once the video is encoded, you can upload it directly to Facebook. If you want to upload to YouTube, exports from AME with the VR box checked should work fine, but for videos from other sources you will need to modify the metadata with this app here.  Once your video is uploaded to YouTube, you can embed it on any webpage that supports 2D web videos. And YouTube videos can be streamed directly to your Rift headset using the free DeoVR video player.

That should give you a 360-video production workflow from start to finish. I will post more updated articles as new software tools are developed, and as I get new 360 cameras with which to test and experiment.

Mike McCarthy is an online editor/workflow consultant with 10 years of experience on feature films and commercials. He has been involved in pioneering new solutions for tapeless workflows, DSLR filmmaking and multi-screen and surround video experiences. Check out his site.

GoPro intros Hero6 and its first integrated 360 solution, Fusion

By Mike McCarthy

Last week, I traveled to San Francisco to attend GoPro’s launch event for its new Hero6 and Fusion cameras. The Hero6 is the next logical step in the company’s iteration of action cameras, increasing the supported frame rates to 4Kp60 and 1080p240, as well as adding integrated image stabilization. The Fusion on the other hand is a totally new product for them, an action-cam for 360-degree video. GoPro has developed a variety of other 360-degree video capture solutions in the past, based on rigs using many of their existing Hero cameras, but Fusion is their first integrated 360-video solution.

While the Hero6 is available immediately for $499, the Fusion is expected to ship in November for $699. While we got to see the Fusion and its footage, most of the hands-on aspects of the launch event revolved around the Hero6. Each of the attendees was provided a Hero6 kit to record the rest of the days events. My group was provided a ride on the RocketBoat through the San Francisco Bay. This adventure took advantage of a number of features of the camera, including the waterproofing, the slow motion and the image stabilization.

The Hero6

The big change within the Hero6 is the inclusion of GoPro’s new custom-designed GP1 image processing chip. This allows them to process and encode higher frame rates, and allows for image stabilization at many frame-rate settings. The camera itself is physically similar to the previous generations, so all of your existing mounts and rigs will still work with it. It is an easy swap out to upgrade the Karma drone with the new camera, which also got a few software improvements. It can now automatically track the controller with the camera to keep the user in the frame while the drone is following or stationary. It can also fly a circuit of 10 waypoints for repeatable shots, and overcoming a limitation I didn’t know existed, it can now look “up.”

There were fewer precise details about the Fusion. It is stated to be able to record a 5.2K video sphere at 30fps and a 3K sphere at 60fps. This is presumably the circumference of the sphere in pixels, and therefore the width of an equi-rectangular output. That would lead us to conclude that the individual fish-eye recording is about 2,600 pixels wide, plus a little overlap for the stitch. (In this article, GoPro’s David Newman details how the company arrives at 5.2K.)

GoPro Fusion for 360

The sensors are slightly laterally offset from one another, allowing the camera to be thinner and decreasing the parallax shift at the side seams, but adding a slight offset at the top and bottom seams. If the camera is oriented upright, those seams are the least important areas in most shots. They also appear to have a good solution for hiding the camera support pole within the stitch, based on the demo footage they were showing. It will be interesting to see what effect the Fusion camera has on the “culture” of 360 video. It is not the first affordable 360-degree camera, but it will definitely bring 360 capture to new places.

A big part of the equation for 360 video is the supporting software and the need to get the footage from the camera to the viewer in a usable way. GoPro already acquired Kolor’s Autopano Video Pro a few years ago to support image stitching for their larger 360 video camera rigs, so certain pieces of the underlying software ecosystem to support 360-video workflow are already in place. The desktop solution for processing the 360 footage will be called Fusion Studio, and is listed as coming soon on their website.

They have a pretty slick demonstration of flat image extraction from the video sphere, which they are marketing as “OverCapture.” This allows a cellphone to pan around the 360 sphere, which is pretty standard these days, but by recording that viewing in realtime they can output standard flat videos from the 360 sphere. This is a much simpler and more intuitive approach to virtual cinematography that trying to control the view with angles and keyframes in a desktop app.

This workflow should result in a very fish-eye flat video, similar to the more traditional GoPro shots, due to the similar lens characteristics. There are a variety of possible approaches to handling the fish-eye look. GoPro’s David Newman was explaining to me some of the solutions he has been working on to re-project GoPro footage into a sphere, to reframe or alter the field of view in a virtual environment. Based on their demo reel, it looks like they also have some interesting tools coming for using the unique functionality that 360 makes available to content creators, using various 360 projections for creative purposes within a flat video.

GoPro Software
On the software front, GoPro has also been developing tools to help its camera users process and share their footage. One of the inherent issues of action-camera footage is that there is basically no trigger discipline. You hit record long before anything happens, and then get back to the camera after the event in question is over. I used to get one-hour roll-outs that had 10 seconds of usable footage within them. The same is true when recording many attempts to do something before one of them succeeds.

Remote control of the recording process has helped with this a bit, but regardless you end up with tons of extra footage that you don’t need. GoPro is working on software tools that use AI and machine learning to sort through your footage and find the best parts automatically. The next logical step is to start cutting together the best shots, which is what Quikstories in their mobile app is beginning to do. As someone who edits video for a living, and is fairly particular and precise, I have a bit of trouble with the idea of using something like that for my videos, but for someone to whom the idea of “video editing” is intimidating, this could be a good place to start. And once the tools get to a point where their output can be trusted, automatically sorting footage could make even very serious editing a bit easier when there is a lot of potential material to get through. In the meantime though, I find their desktop tool Quik to be too limiting for my needs and will continue to use Premiere to edit my GoPro footage, which is the response I believe they expect of any professional user.

There are also a variety of new camera mount options available, including small extendable tripod handles in two lengths, as well as a unique “Bite Mount” (pictured, left) for POV shots. It includes a colorful padded float in case it pops out of your mouth while shooting in the water. The tripods are extra important for the forthcoming Fusion, to support the camera with minimal obstruction of the shot. And I wouldn’t recommend the using Fusion on the Bite Mount, unless you want a lot of head in the shot.

Ease of Use
Ironically, as someone who has processed and edited hundreds of hours of GoPro footage, and even worked for GoPro for a week on paper (as an NAB demo artist for Cineform during their acquisition), I don’t think I had ever actually used a GoPro camera. The fact that at this event we were all handed new cameras with zero instructions and expected to go out and shoot is a testament to how confident GoPro is that their products are easy to use. I didn’t have any difficulty with it, but the engineer within me wanted to know the details of the settings I was adjusting. Bouncing around with water hitting you in the face is not the best environment for learning how to do new things, but I was able to use pretty much every feature the camera had to offer during that ride with no prior experience. (Obviously I have extensive experience with video, just not with GoPro usage.) And I was pretty happy with the results. Now I want to take it sailing, skiing and other such places, just like a “normal” GoPro user.

I have pieced together a quick highlight video of the various features of the Hero6:

Mike McCarthy is an online editor/workflow consultant with 10 years of experience on feature films and commercials. He has been involved in pioneering new solutions for tapeless workflows, DSLR filmmaking and multi-screen and surround video experiences. Check out his site.

Dell 6.15

Making the jump to 360 Video (Part 1)

By Mike McCarthy

VR headsets have been available for over a year now, and more content is constantly being developed for them. We should expect that rate to increase as new headset models are being released from established technology companies, prompted in part by the new VR features expected in Microsoft’s next update to Windows 10. As the potential customer base increases, the software continues to mature, and the content offerings broaden. And with the advances in graphics processing technology, we are finally getting to a point where it is feasible to edit videos in VR, on a laptop.

While a full VR experience requires true 3D content, in order to render a custom perspective based on the position of the viewer’s head, there is a “video” version of VR, which is called 360 Video. The difference between “Full VR” and “360 Video,” is that while both allow you to look around every direction, 360 Video is pre-recorded from a particular point, and you are limited to the view from that spot. You can’t move your head to see around behind something, like you can in true VR. But 360 video can still offer a very immersive experience and arguably better visuals, since they aren’t being rendered on the fly. 360 video can be recorded in stereoscopic or flat, depending on the capabilities of the cameras used.

Stereoscopic is obviously more immersive, less of a video dome and inherently supported by the nature of VR HMDs (Head Mounted Displays). I expect that stereoscopic content will be much more popular in 360 Video than it ever was for flat screen content. Basically the viewer is already wearing the 3D glasses, so there is no downside, besides needing twice as much source imagery to work with, similar to flat screen stereoscopic.

There are a variety of options for recording 360 video, from a single ultra-wide fisheye lens on the Fly360, to dual 180-degree lens options like the Gear 360, Nikon KeyMission, and Garmin Virb. GoPro is releasing the Fusion, which will fall into this category as well. The next step is more lens, with cameras like the Orah4i or the Insta360 Pro. Beyond that, you are stepping into the much more expensive rigs with lots of lenses and lots of stitching, but usually much higher final image quality, like the GoPro Omni or the Nokia Ozo. There are also countless rigs that use an array of standard cameras to capture 360 degrees, but these solutions are much less integrated than the all-in-one products that are now entering the market. Regardless of the camera you use, you are going to be recording one or more files in a pixel format fairly unique to that camera that will need to be processed before it can be used in the later stages of the post workflow.

Affordable cameras

The simplest and cheapest 360 camera option I have found is the Samsung Gear 360. There are two totally different models with the same name, usually differentiated by the year of their release. I am using the older 2016 model, which has a higher resolution sensor, but records UHD instead of the slightly larger full 4K video of the newer 2017 model.

The Gear 360 records two fisheye views that are just over 180 degrees, from cameras situated back to back in a 2.5-inch sphere. Both captured image circles are recorded onto a single frame, side by side, resulting in a 2:1 aspect ratio files. These are encoded into JPEG (7776×3888 stills) or HEVC (3840×1920 video) at 30Mb and saved onto a MicroSD card. The camera is remarkably simple to use, with only three buttons, and a tiny UI screen to select recording mode and resolution. If you have a Samsung Galaxy phone, there are a variety of other functions that allows, like remote control and streaming the output to the phone as a viewfinder and such. Even without a Galaxy phone, the camera did everything I needed to generate 360 footage to stitch and edit with but it was cool to have a remote viewfinder for the driving shots.

Pricier cameras

One of the big challenges of shooting with any 360 camera is how to avoid getting gear and rigging in the shot since the camera records everything around it. Even the tiny integrated tripod on the Gear 360 is visible in the shots, and putting it on the plate of my regular DSLR tripod fills the bottom of the footage. My solution was to use the thinnest support I could to keep the rest of the rigging as far from the camera as possible, and therefore smaller from its perspective. I created a couple options to shoot with that are pictured below. The results are much less intrusive in the resulting images that are recorded. Obviously besides the camera support, there is the issue of everything else in the shot including the operator. Since most 360 videos are locked off, an operator may not be needed, but there is no “behind the camera” for hiding gear or anything else. Your set needs to be considered in every direction, since it will all be visible to your viewer. If you can see the camera, it can see you.

There are many different approaches to storing 360 images, which are inherently spherical, as a video file, which is inherently flat. This is the same issue that cartographers have faced for hundreds of years — creating flat paper maps of a planet that is inherently curved. While there are sphere map, cube map and pyramid projection options (among others) based on the way VR headsets work, the equirectangular format has emerged as the standard for editing and distribution encoding, while other projections are occasionally used for certain effects processing or other playback options.

Usually the objective of the stitching process is to get the images from all of your lenses combined into a single frame with the least amount of distortion and the fewest visible seams. There are a number of software solutions that do this, from After Effects plugins, to dedicated stitching applications like Kolor AVP and Orah VideoStitch-Studio to unique utilities for certain cameras. Once you have your 360 video footage in the equirectangular format, most of the other steps of the workflow are similar to their flat counterparts, besides VFX. You can cut, fade, title and mix your footage in an NLE and then encode it in the standard H.264 or H.265 formats with a few changes to the metadata.

Technically, the only thing you need to add to an existing 4K editing workflow in order to make the jump to 360 video is a 360 camera. Everything else could be done in software, but the other thing you will want is a VR headset or HMD. It is possible to edit 360 video without an HMD, but it is a lot like grading a film using scopes but no monitor. The data and tools you need are all right there, but without being able to see the results, you can’t be confident of what the final product will be like. You can scroll around the 360 video in the view window, or see the whole projected image all distorted, but it won’t have the same feel as experiencing it in a VR headset.

360 Video is not as processing intensive as true 3D VR, but it still requires a substantial amount of power to provide a good editing experience. I am using a Thinkpad P71 with an Nvidia Quadro P5000 GPU to get smooth performance during all these tests.

Stay tuned for Part 2 where we focus on editing 360 Video.

Mike McCarthy is an online editor/workflow consultant with 10 years of experience on feature films and commercials. He has been working on new solutions for tapeless workflows, DSLR filmmaking and multi-screen and surround video experiences. Check out his site.

Director Ava DuVernay named VES Summit’s keynote speaker

Director/producer/writer Ava DuVernay has been named keynote speaker at the 2017 VES Summit, “Inspiring Change: Building on 20 Years of VES Innovation.” The forum, which takes place Saturday, October 28, celebrates the Visual Effects Society’s 20th anniversary and brings together creatives, executives and visionaries from a variety of disciplines to discuss the evolution of visual imagery and the VFX industry landscape in a TED Talks-like atmosphere.

At the 2012 Sundance Film Festival, DuVernay won the Best Director Prize for her second feature film Middle of Nowhere, which she also wrote and produced. For her work on Selma in 2014, she was nominated for the Academy Award for Best Picture. In 2017, she was nominated for the Academy Award for Best Documentary Feature for her film 13th. Her current directorial work includes the dramatic television series Queen Sugar, and the upcoming Disney feature film A Wrinkle in Time.

It was back in 2010 that DuVernay made her directorial debut with the acclaimed 2008 hip-hop documentary This Is The Life, and she has gone on to direct several network documentaries, including Venus Vs. for ESPN. She has also directed significant short form work, including August 28: A Day in the Life of a People, commissioned by The Smithsonian’s National Museum of African American History and Culture, as well as fashion and beauty films for Prada and Apple.

Other speakers include:
–  Syd Mead, visual futurist and conceptual artist
–  President of IMAX Home Entertainment Jason Brenek on “Evolution in Entertainment: VR, Cinema and Beyond”
– CEO of SSP BlueHemanshu Nigam on “When Hackers Attack: How Can Hollywood Fight Back?”
– Head of Adobe Research Gavin Miller on “Will the Future Look More Like Harry Potter or Star Trek?”
–  Senior research engineer at Autodesk, Evan Atherton on “The Age of Imagination”
–  Founder/CEO of the Emblematic Group, Nonny de la Peña on “Creating for Virtual, Augmented and Mixed Realities”

Additional speakers and roundtable moderators will be announced soon. The 2017 VES Summit takes place at the Sofitel Hotel Beverly Hills.

Review: Blackmagic’s Fusion 9

By David Cox

At Siggraph in August, Blackmagic Design released a new version of its compositing software Fusion. For those not familiar with Fusion, it is a highly flexible node-based compositor that can composite in 2D and 3D spaces. Its closest competitor is Nuke from The Foundry.

The raft of new updates in Version 9 could be categorized into one of two areas: features created in response to user requests, and a set of tools for VR. Also announced with the new release is a price drop to $299 for the full studio version, which, judging by global resellers instantly running out of stock (Fusion ships via dongle), seems to have been a popular move!

As with other manufacturers in the film and broadcast area, the term “VR” is a little misused as they are really referring to “360 video.” VR, although a more exciting term, would demand interactivity. That said, as a post production suite for 360 video, Fusion already has a very strong tool set. It can create, manipulate, texture and light 3D scenes made from imported CGI models and built-in primitives and particles.

Added in Version 9 is a spherical camera that can capture a scene as a 360 2D or stereo 3D image. In addition, new tools are provided to cross-convert between many 360 video image formats. Another useful tool allows a portion of a 360-degree image to be unwrapped (or un-distorted) so that restoration or compositing work can be easily carried out on it before it is perfectly re-wrapped back into the 360-degree image.

There is also a new stabilizer for 360 wrap-around shots. A neat feature is that Fusion 9 can directly drive VR headsets such as Oculus Rift. Within Fusion, any node can be routed to any viewing monitor and the VR headset simply presents itself as an extra one of those.

Notably, Blackmagic has opted not to tackle 360-degree image stitching — the process by which images from multiple cameras facing in different directions are “stitched” together to form a single wrap-around view. I can understand this — on one hand, there are numerous free or cheap apps that perform stitching and so there’s no need for Blackmagic to reinvent that wheel. On the other hand, Blackmagic targets the mass user area, and given that 360 video production is a niche activity, productions that strap together multiple cameras form an even smaller and decreasing niche due to the growing number of single-step 360-degree cameras that provide complete wrap-around images without the need for stitching.

Moving on from VR/360, Fusion 9 now boasts some very significant additional features. While some Fusion users had expressed concerned that Blackmagic was favoring Resolve, in fact it is now clear that the Fusion development team have been very busy indeed.

Camera Tracker
First up is an embedded camera tracker and solver. Such a facility aims to deduce how the original camera in a live-action shoot moved through the scene and what lens must have been on it. From this, a camera tracker produces a virtual 3D scene into which a compositor can add objects that then move precisely with the original shot.

Fusion 9’s new camera tracker performed well in tests. It requires the user to break the process down into three logical steps: track, refine and export. Fusion initially offers auto-placed trackers, which follow scores of details in the scene quite quickly. The operator then removes any obviously silly trackers (like the ones chasing around the moving people in a scene) and sets Fusion about the task of “solving” the camera move.

Once done, Fusion presents a number of features to allow the user to measure the accuracy of the resulting track and to locate and remove trackers that are adversely affecting that result. This is a circular process by which the user can incrementally improve the track. The final track is then converted into a 3D scene with a virtual camera and a point cloud to show where the trackers would exist in 3D space. A ground plane is also provided, which the user can locate during the tracking process.

While Fusion 9’s camera tracker perhaps doesn’t have all the features of a dedicated 3D tracker such as SynthEyes from Andersson Technologies, it does satisfy the core need and has plenty of controls to ensure that the tool is flexible enough to deal with most scenarios. It will certainly be received as a welcome addition.

Planar Tracker
Next up is a built-in “planar” tracker. Planar trackers work differently than classic point trackers, which simply try to follow a small area of detail. A planar tracker follows a larger area of a shot, which makes up a flat plane — such as a wall or table top. From this, the planar tracker can deduce rotation, location, scale and perspective.

Fusion 9 Studio’s new planar tracker also performed well in tests. It assessed the track quickly and was not easily upset by foreground objects obscuring parts of the tracked area. The resulting track can either be used directly to insert another image into the resulting plane or to stabilize the shot, or indirectly by producing a separate Planar Transform node. This is used to warp any other asset such as a matte for rotoscoping work.

Inevitably, any planar tracker will be compared to the long-established “daddy” of them all, Mocha Pro from Boris FX. At a basic level, Fusion’s planar tracker worked just as well as Mocha, creating solid tracks from a user-defined area nicely and quickly. However, I would think that for complex rotoscoping, where a user will have many roto layers, driven by many tracking sources, with other layers acting as occlusion masks, Mocha’s working environment would be easier to control. Such a task would lead to many, many wired up nodes in Fusion, whereas Mocha would present the same functions within a simper layer-list. Of course, Mocha Pro is available as an OFX plug-in for Fusion Studio anyway, so users can have the best of both worlds.

Delta Keyer
Blackmagic also added a new keyer to Fusion called the Delta Keyer. It is a color difference keyer with a wide range of controls to refine the resulting matte and the edges of the key. It worked well when tested against one of my horrible greenscreens, something I keep for these very occasions!

The Delta Keyer can also take a clean plate as a reference input, which is essentially a frame of the green/bluescreen studio without the object to be keyed. The Delta Keyer then uses this to understand which deviations from the screen color represent the foreground object and which are just part of an uneven screen color.

To assist with this process, there is also a new Clean Plate node, which is designed to create an estimate of a clean plate in the absence of one being available from the shoot (for example, if the camera was moving). The combination of the clean plate and the Delta Keyer produced good results when challenged to extract subtle object shadows from an unevenly lit greenscreen shot.

Studio Player
Studio Player is also new for Fusion 9 Studio; it’s a multi-station shot review tool. Multiple versions of clips and comps can be added to the Studio Player’s single layer timeline, where simple color adjustments and notes can be added. A neat feature is that multiple studio players in different locations can be slaved together so that cross-facility review sessions can take place, with everyone looking at the same thing at the same time, which helps!

Fusion 9 Studio also supports the writing of Apple-approved Pro Res from all its supported platforms, including Windows and Linux. Yep – you read that right. Other format support has also been widened and improved, such as faster native handling for DNxHR codecs, for example.

Summing Up
All in all, the updates to Fusion 9 are comprehensive and very much in line with what professional users have been asking for. I think it certainly demonstrates that Blackmagic is as committed to Fusion as Resolve, and at $299, it’s a no-brainer for any professional VFX artist to have available to them.

Of course, the price drop shows that Blackmagic is also aiming Fusion squarely at the mass independent filmmaker market. Certainly, with Resolve and Fusion, those users will have pretty much all the post tools they will need.

Fusion by its nature and heritage is a more complex beast to learn than Resolve, but it is well supported with a good user manual, forums and video tutorials. I would think it likely that for this market, Fusion might benefit from some minor tweaks to make it more intuitive in certain areas. I also think the join between Resolve and Fusion will provide a lot of interest going forward for this market. Adobe has done a masterful job bridging Premiere and After Effects. The join between Resolve and Fusion is more rudimentary, but if Blackmagic gets this right, they will have a killer combination.

Finally, Fusion 9 extends what was already a very powerful and comprehensive compositing suite. It has become my primary compositing device and the additions in version 9 only serve to cement that position.

David Cox is a VFX compositor and colorist with 20+ years experience. He started his career with MPC and The Mill before forming his own London-based post facility. Cox recently created interactive projects with full body motion sensors and 4D/AR experiences.

Charlieuniformtango adds director Elliot Dillman to roster

Director Elliot Dillman has joined Charlieuniformtango for national commercial and VR representation. His directorial experience spans broadcast commercials, branded VR content, music videos and multi-camera live events.

He has been at the helm of national ad campaigns for a number of top agencies including GSD&M, CP+B, Leo Burnett, Y&R and BBDO, directing spots for brands such as Kraft, Nerf, GMC and Subway.

Recently, Dillman’s work with Verizon and Momentum Worldwide received two 2017 Clio Awards for the “Virtual Gridiron” VR experience at Super Bowl LI. Also, his short film, “In Harmony,” part of the Oculus VR for Good program, premiered at the Oculus house during the 2017 Sundance Film Festival and was an official selection of SXSW 2017. The film examines the Harmony Project’s work to help Los Angeles kids stay in school through educational music programs. “

Dillman is the oldest son of Emmy-winning director Ray Dillman, so he grew up on film sets. He started working regularly as a PA at age 12 for production companies like Gartner and MJZ. While studying at Loyola Marymount University’s School of Film and Television, Dillman simultaneously began his directing career with multi-camera live concert shoots for the ESPN Summer and Winter X Games. Shortly after graduating, he directed feature film tie-in spots for Sony Pictures, Warner Bros. and Paramount, along with ad campaigns for the Walt Disney Company and Royal Purple Motor Oil, amongst others.

Review: Lenovo’s ThinkPad P71 mobile workstation

By Mike McCarthy

Lenovo was nice enough to send me their newest VR-enabled mobile workstation to test out on a VR workflow project I am doing. The new ThinkPad P71 is a beast with a 17-inch UHD IPS screen. The model they sent to me was equipped with the fastest available processor, a Xeon E5-1535M v6 with four cores processing eight threads at an official speed of 3.1GHz. It has 32GB of DDR4-2400 ECC RAM, with two more slots allowing that to be doubled to 64GB if desired.

The system’s headline feature is the Nvidia Quadro P5000 mobile GPU, with 2,048 CUDA cores, fed by another 16GB of dedicated DDR5 memory. The storage configuration is a single NVMe 1TB SSD populating one of two available M.2 slots. This configuration is currently available for $5,279, discounted to $4,223.20 on right now. So while it is not cheap, it is one of the most powerful mobile workstations you can buy right now.

Connectivity wise, it has dual Thunderbolt 3 ports, which can also be used for USB 3.1 Type C devices. It has four more USB 3.1 Type A ports and a Gigabit Ethernet port. You have a number of options for display connectivity. Besides the Thunderbolt ports, there is a MiniDP 1.2 port and an HDMI 1.4 port (1.4 based on Intel graphics limitations). It has an SDXC slot, an ExpressCard34 slot, and a single 1/8-inch headphone mic combo jack. The system also has a docking connector and a rectangular port for the included 230W power adaptor.

It has the look and feel of a traditional ThinkPad, which goes back to the days when they were made by IBM. It has the customary TrackPoint as well as a touchpad. Both have three mouse buttons, which I like in theory, but I constantly find myself trying to click with the center button to no avail. I would either have to get used to it, or set the center action to click as well, defeating the purpose of the third button. The Fn key in the bottom corner will take some getting used to as well, as I keep hitting that instead of CTRL, but I adapted to a similar configuration on my current laptop.

I didn’t like the combo jack at first, because it required a cheap adapter, but now that I have gotten one, I see why that is the future, once all the peripherals support it. I had plugged my mic and headphones in backwards as recently as last week, so it is an issue when the ports aren’t clearly labeled and the combo jack solves that once and for all. It is a similar jack to most cell phones, and you only need an adapter for the mic functionality, regular headphones work by default.

The system doesn’t weigh as much as I expected, probably due to the lack of spinning disks or optical drive, which can be added if desired. It came relatively clean, with Windows 10 Pro installed, without too many other applications or utilities pre-installed. It had all of the needed drivers and a simple utility for operating the integrated X-Rite Pantone color calibrator for the screen. There was a utility for adding any other applications that would normally be included, which I used to download the Lenovo Performance Tuner. I use the Performance Tuner more for monitoring usage than adjusting settings, but can be nice to have everything in one place, especially in Windows 10.

The system boots up in about 10 seconds, and shuts down even faster. Hibernating takes twice as long, which is to be expected with that much RAM to be cached to disk, even with an NVMe SSD. But that may be worth the extra time to keep your applications open. My initial tests of the SSD showed a 1700MB/s write speed with 2500MB/s reads. Longer endurance tests resulted in write speeds decreasing to 1200MB/s, but the read speeds remained consistently above 2500MB/s. That should be more than enough throughput for most media work, even allowing me to playback uncompressed 6K content, and should allow 4K uncompressed media capture if you connect an I/O device to the Thunderbolt bus.

The main application I use on a daily basis is Adobe Premiere Pro, so most of my performance evaluation revolves around that program, although I used a few others as well. I was able to load a full feature film off of a USB3 drive with no issues. The 6K Cineform and DNxHR media played back at ½ res without issue. The 6K R3D files played at ¼ res without dropping frames, which is comparable to my big tower.

My next playback test was fairly unique to my workflow, but a good benchmark of what is possible. I was able to connect three 1080p televisions to the MiniDP port, using an MST (Multi-Stream Transport) hub, with three HDMI ports. Using the Nvidia Mosaic functionality offered by the Quadro P5000 card, I can span them into a single display, which Premiere can send output to, via the Adobe’s Mercury Playback engine. This configuration allows me to playback 6K DNxHR 444 files to all three screens, directly off the timeline, at half res, without dropping frames. My 6K H.265 files playback at full res outside Premiere. That is a pretty impressive display for a laptop. Once I had maxed out the possibilities for playback, I measured a few encodes. In general, the P71 takes about twice as long to encode things in Adobe Media Encoder as my 20-core desktop workstation, but is twice as fast as my existing quad Core i7 4860 laptop.

The other application I have been taxing my system with recently is DCP-O-Matic. It takes 30 hours to render my current movie to a 4K DCP on my desktop, which is 18x the runtime, but I know most of my system’s 20 cores are sitting idle based on the software threading. Doing a similar encode on the Lenovo system took 12.5x the run time, so that means my 100-minute film should take 21 hours. The higher base frequency of the quad core CPU definitely makes a difference in this instance.

The next step was to try my HMD headset with it to test out the VR capability. My Oculus Rift installed without issues, which is saying something, based on the peculiarities of Oculus’ software. Maybe there is something to that “VR-ready” program, but I did frequently have issues booting up the system with the Rift connected, so I recommend plugging it in after you have your system up and running. Everything VR-related ran great, except for the one thing I actually wanted to do, which was edit 360 video in Premiere, with the HMD. There was some incompatibility between the drivers for the laptop and the software.

There are a variety of ways to test battery life, but since this is a VR-ready system that seemed to be the best approach. How long would it support using a VR headset before needing to plug in? I got just short of an hour of heavy 3D VR usage before I started getting low battery warnings. I was hoping to be able to close the display to save power, since I am not looking at it while using the headset. (I usually set the Close Lid action to Do Nothing on all my systems because I want to be able to walk into the other room to show someone something on my timeline without effecting the application. If I want to sleep the system, I can press the button.) But whenever the Rift is active, closing the lid puts the machine to sleep immediately, regardless of the settings. So you have to run the display and the HMD anytime you are working in VR. And don’t plan on doing extensive work without plugging in.

Now to be fair, setting up to use VR involves preparing the environment and configuring sensors, so adding power to that mix is a reasonable requirement and very similar to 3D gaming. Portable doesn’t always mean untethered. But for browsing the Internet, downloading project files and editing articles, I would expect about four hours of battery life from the system before needing to recharge. It is really hard to accurately estimate run time when the system’s performance and power needs scale so much depending on the user’s activities. The GPU alone scales from 5 watts to 100 watts depending on what is being processed, but the run time is not out of line with what is to be expected from products in this class of performance.

Summing Up
All in all, the P71 is an impressive piece of equipment, and one of only a few ways you can currently get a portable professional VR solution. I recognize that most of my applications aren’t using all of the power I would be carrying around in a P71, so for my own work, I would probably hope to find a smaller and lighter-weight system at the expense of some of that processing power. But for people who have uncompromising needs for the fastest system they can possibly get, the Lenovo P71 fits the bill. It is a solid performer that can do an impressive amount of processing, while still being able to come with you wherever you need to go.

Mike McCarthy is an online editor/workflow consultant with 10 years of experience on feature films and commercials. He has been working on new solutions for tapeless workflows, DSLR filmmaking and multi-screen and surround video experiences. Check out his site.


postPerspective Impact Award winners from SIGGRAPH 2017

Last April, postPerspective announced the debut of our Impact Awards, celebrating innovative products and technologies for the post production and production industries that will influence the way people work. We are now happy to present our second set of Impact Awards, celebrating the outstanding offerings presented at SIGGRAPH 2017.

Now that the show is over, and our panel of VFX/VR/post pro judges has had time to decompress, dig out and think about what impressed them, we are happy to announce our honorees.

And the winners of the postPerspective Impact Award from SIGGRAPH 2017 are:

  • Faceware Technologies for Faceware Live 2.5
  • Maxon for Cinema 4D R19
  • Nvidia for OptiX 5.0  

“All three of these technologies are very worthy recipients of our first postPerspective Impact Awards from SIGGRAPH,” said Randi Altman, postPerspective’s founder and editor-in-chief. “These awards celebrate companies that define the leading-edge of technology while producing tools that actually make users’ working lives easier and projects better, and our winners certainly fall into that category.

“While SIGGRAPH’s focus is on VFX, animation, VR/AR and the like, the types of gear they have on display vary. Some are suited for graphics and animation, while others have uses that slide into post production. We’ve tapped real-world users in these areas to vote for our Impact Awards, and they have determined what tools might be most impactful to their day-to-day work. That’s what makes our awards so special.”

There were many new technologies and products at SIGGRAPH this year, and while only three won an Impact Award, our judges felt there were other updates that it was important to let people know about as well.

Blackmagic Design’s Fusion 9 was certainly turning heads and Nvidia’s VRWorks 360 Video was called out as well. Chaos Group also caught our judges attention with V-Ray for Unreal Engine 4.

Stay tuned for future Impact Award winners in the coming months — voted on by users for users — from IBC.

Quick Look: Jaunt One’s 360 camera

By Claudio Santos

To those who have been following the virtual reality market from the beginning, one very interesting phenomenon is how the hardware development seems to have outpaced both the content creation and the software development. The industry has been in a constant state of excitement over the release of new and improved hardware that pushes the capabilities of the medium, and content creators are still scrambling to experiment and learn how to use the new technologies.

One of the products of this tech boom is the Jaunt One camera. It is a 360 camera that was developed with the explicit focus of addressing the many production complexities that plague real life field shooting. What do I mean by that? Well, the camera quickly disassembles and allows you to replace a broken camera module. After all, when you’re across the world and the elephant that is standing in your shot decides to play with the camera, it is quite useful to be able to quickly swap parts instead of having to replace the whole camera or sending it in for repair from the middle of the jungle.

Another of the main selling points of the Jaunt One camera is the streamlined cloud finishing service they provide. It takes the content creator all the way from shooting on set through stitching, editing, onlining and preparing the different deliverables for all the different publishing platforms available. The pipeline is also flexible enough to allow you to bring your footage in and out of the service at any point so you can pick and choose what services you want to use. You could, for example, do your own stitching in Nuke, AVP or any other software and use the Jaunt cloud service to edit and online these stitched videos.

The Jaunt One camera takes a few important details into consideration, such as the synchronization of all of the shutters in the lenses. This prevents stitching abnormalities in fast moving objects that are captured in different moments in time by adjacent lenses.

The camera doesn’t have an internal ambisonics microphone but the cloud service supports ambisonic recordings made in a dual system or Dolby Atmos. It was interesting to notice that one of the toolset apps they released was the Jaunt Slate, a tool that allows for easy slating on all the cameras (without having to run around the camera like a child, clapping repeatedly) and is meant to automatize the synchronization of the separate audio recordings in post.

The Jaunt One camera shows that the market is maturing past its initial DIY stage and the demand for reliable, robust solutions for higher budget productions is now significant enough to attract developers such as Jaunt. Let’s hope tools such as these encourage more and more filmmakers to produce new content in VR.

Blackmagic’s Fusion 9 is now VR-enabled

At SIGGRAPH, Blackmagic was showing Fusion 9, its newly upgraded visual effects, compositing, 3D and motion graphics software. Fusion 9 features new VR tools, an entirely new keyer technology, planar tracking, camera tracking, multi-user collaboration tools and more.

Fusion 9 is available now with a new price point — Blackmagic has lowered the price of its Studio version from $995 to $299 Studio Version. (Blackmagic is also offering a free version of Fusion.) The software now works on Mac, PC and Linux.

Those working in VR get a full 360º true 3D workspace, along with a new panoramic viewer and support for popular VR headsets such as Oculus Rift and HTC Vive. Working in VR with Fusion is completely interactive. GPU acceleration makes it extremely fast so customers can wear a headset and interact with elements in a VR scene in realtime. Fusion 9 also supports stereoscopic VR. In addition, the new 360º spherical camera renders out complete VR scenes, all in a single pass and without the need for complex camera rigs.

The new planar tracker in Fusion 9 calculates motion planes for accurately compositing elements onto moving objects in a scene. For example, the new planar tracker can be used to replace signs or other flat objects as they move through a scene. Planar tracking data can also be used on rotoscope shapes. That means users don’t have to manually animate motion, perspective, position, scale or rotation of rotoscoped elements as the image changes.

Fusion 9 also features an entirely new camera tracker that analyzes the motion of a live-action camera in a scene and reconstructs the identical motion path in 3D space for use with cameras inside of Fusion. This lets users composite elements with precisely matched movement and perspective of the original. Fusion can also use lens metadata for proper framing, focal length and more.

The software’s new delta keyer features a complete set of matte finesse controls for creating clean keys while preserving fine image detail. There’s also a new clean plate tool that can smooth out subtle color variations on blue- and greenscreens in live action footage, making them easier to key.

For multi-user collaboration, Fusion 9 Studio includes Studio Player, a new app that features a playlist,
storyboard and timeline for playing back shots. Studio Player can track version history, display annotation notes, has support for LUTs and more. The new Studio Player is suited for customers that need to see shots in a suite or theater for review and approval. Remote synchronization lets artists  sync Studio Players in multiple locations.

In addition, Fusion 9 features a bin server so shared assets and tools don’t have to be copied onto each user’s local workstation.