generative AI
People who are acquainted with Faizan Buzdar, who formerly served as the vice president of product management at Box, are probably familiar with Convo, a popular digital workplace platform used by newsrooms like this one. However, Buzdar, an electronics engineer by training, has always been fascinated with film and visual effects.
As a lifetime fan of both video and photography, Buzdar said to me through email, “I’d been making videos on my own for years but noticed video production had largely remained manual with little innovation in recent decades, especially for time-consuming tasks like video editing.” Meanwhile, I observed that over the last several years, iPhone camera and sensor technology had made significant advancements, nearly matching those of DSLRs in terms of image quality.
Buzdar claims that while working at Box, he made the decision to experiment with fusing video—a rapidly growing medium—with advancements in AI and machine learning in an effort to enhance the experience of recording and editing video. Together with Saj Khan, Fahad Yaqub, and fellow Box executive Michelle Oh, Buzdar recruited video engineer Adeel Abbas, who while working at Twitter helped build the foundation supporting the site’s livestreaming capabilities. The group is exploring the possibilities of tech-accelerated video creation.
The result is a roll. It offers motion graphics, bokeh, multicam shots, and — probably most intriguingly to me — “AI-simulated” sliders, dollies, and jibs. It is a new iOS app.
Our goal is to upend the high-quality video production industry and establish ourselves as the new benchmark for producing video content, said Buzdar. We’re doing away with all of that because producing outstanding video requires a significant upfront investment in tools, equipment, training, and editing software.
Roll comprises of two products: the Roll iPhone app and the Roll web app. Roll is targeted at the “prosumer” market (think influencers and podcasters, but also companies developing their own marketing content). The iPhone software records video, captures it, and instantly transfers it to Roll’s cloud for processing and storage. While this is going on, content makers may preview, access, share, download, and edit footage via the web app.
Of course, there are a tonne of video applications available. So what distinguishes Roll? One of the reasons, according to Buzdar, is that the app is targeted at use cases that other camera applications aren’t: distant video interviews, video podcasts, and client testimonials. Buzdar contends that although Zoom, Microsoft Teams, and Google Meet partially meet the demand, they are not intended for “high-quality” video creation.
Roll additionally makes use of a few real-time effects to (ostensibly) provide more post-production options than the majority of video-capturing apps. By using the HEVC standard, for instance, Roll may produce a bitrate that is nearly twice as high and a picture quality that is greater for a given file size. A wide-angle view and a close-up shot may be recorded and processed by And Roll at the same time, enabling users to make films that essentially have “multi-camera” perspectives.
Granted, multicam isn’t particularly unique — roll’s far from the first app to offer it. But Buzdar says that where the magic lies is in the post-processing. Roll leverages generative AI to recreate rooms in 3D space so that content creators can move a video-game-like virtual camera around, simulating movements like panning from side to side with a dolly or crane.
“Today, generative AI is too often associated with creating fake content out of thin air,” Buzdar said. “That’s not our philosophy. We don’t generate fake pixels, people or scenes. We’re using generative AI purely as a tool for productivity — we want to democratize access to higher quality video production.”
According to Buzdar, Roll’s AI was trained to recognise the 3D depth in a scene by using data to assess depth and forms separate from the presence of a human observer in the space. Roll first used open source datasets that are often used for benchmarking in academia to train its algorithms, but later internally recorded over 22,000 video conversations, building its own rich database.
At least in the demo clip that Buzdar showed me, the results aren’t too awful. Some of Roll’s AI-generated pans, the consequence of weird bending on backdrop objects as the virtual camera swivels by, cross the uncanny valley. However, in brief sequences, the AI effects are sufficient and bring interest to what would otherwise be a tedious distant interview.
According to Buzdar, “We’ve done a lot of research on this and we haven’t seen anyone use AI in the same way we are — pairing iPhone sensor data with large AI models in the cloud.” “Our technology offers the fundamental capabilities to simulate any desired visual effects,”
Any visual effects seem a bit improbable. Roll, however, is equipped with additional, more realistic computational trickery. Roll collects information on the recording and lighting circumstances, the subject’s location, the distance between the camera and the subject, and other details as it captures video. The information is utilised to offer feedback and instructions for composition and lighting as well as automatically adjusting the phone’s cameras and sensors.
Roll uses the information, like a few other “AI-enabled” mobile video editors on the market, to build a fully realised, multicamera reel in its editing cloud without the need for manual editing. (Users still have the option to modify the camera angles, add camera motions, and add graphic effects.) Roll will soon be able to upload content straight to social media platforms like TikTok, YouTube, and Instagram in the proper quality and aspect ratio.
According to Buzdar, “today’s video production requires numerous pieces of hardware and software to be completed.” “With every single step, the video and audio file loses meaning and basically turns into a ‘dumb’ file that is handed around when it travels from one software to another. The whole “stack” of video creation has been completely reconstructed by us. We’ve deployed AI to provide a revolutionary capture-to-publish experience that vertically integrates and automates the whole remote video production workflow, transcending traditional software boundaries.
So, how does Roll intend to monetize? The business has so far received funding from conventional VC sources, though Buzdar declined to specify which ones. Buzdar thinks that Roll will ultimately expand to meet the demands of corporate organisations, particularly their internal corporate marketing and video teams, who would pay a charge for Roll’s services. This would be in terms of income generating.
The cloud is ready to disrupt the video production industry, according to Buzdar. It is the ideal candidate for experiencing the exponential benefits of cloud computing, including scalable storage, AI, computation, and real-time sharing and collaboration because of characteristics like enormous file volumes, complicated processing, and the requirement for multiperson revisions and review cycles.
That is undoubtedly true. Time will tell if Roll turns out to be the disruptor.
