Visual Data Takes a Quantum Leap into UHD and AI-Powered Post Production

Published on in Miscellaneous

Visual Data Media Services provides post-production services, film scanning and transcoding for a wide range of international customers, preparing content for distribution across the world’s premiere broadcast channels and digital platforms. For more than two decades, the company has successfully provided everything from ingest and scanning to editing, captioning and localization, growing to become a premier resource for the entire mastering and delivery process. But last year, with a tremendous increase in 4K and ultra high definition (UHD) projects, Visual Data’s systems were struggling to keep up.

“Whenever we had to process uncompressed 4K files, the existing NAS infrastructure was just too slow to handle more than one job at a time,” said Visual Data Director of Engineering Ken Spickler. “The team got the work out, but managing around the bottleneck was a major problem.”

So Visual Data began to look for a storage solution that could double capacity and deliver the performance to support multiple 4K and UHD operations at the same time.

But capacity to support 4K workloads was just one part of the equation; Visual Data also wanted to take advantage of a major new opportunity to solve the problem of quickly and efficiently re-mastering existing content. During the 90s and early 2000s, when the first non-linear editing systems were used for episodic TV and direct-to-video motion pictures, work was shot on film initially, then production teams edited and finished on standard-definition video, eliminating the need to create a cut negative of the final version. That new workflow made everything faster and less expensive, but it created a nightmare when content owners wanted to distribute their standard definition (SD) content in high definition (HD) or 4K.

With no cut negative to scan, the only way to get old features to 4K was to perform a match-back— scanning the original dailies and eye matching against the finished SD element in order to conform the content. Traditional match-back is incredibly slow and expensive because editors have to search through all shot film, one scene at a time, and match them with the original images in the dailies. It takes weeks just to find the images in a single, 30-minute TV series episode, so it can take years—and millions of dollars—to process multiple seasons of an hour-long program. The company needed a better re-mastering solution.

Late in 2017, after more than a year of internal development, Visual Data was ready to offer its revolutionary Matchmaker match-back service, leveraging image-matching software that uses artificial intelligence (AI) to automate the process. With Matchmaker, Visual Data scans the dailies in 2K or 4K, ingests a proxy of the finished SD content for reference, runs the image detection software to automatically identify the needed frames for each scene, and creates an edit decision list that lands in the Resolve color grading application. Then, Visual Data finishes the process by restoring the picture and audio, color correcting and transcoding out to any required deliverable.

“Right now, the application eliminates more than 95% of the manual work involved in the match-back process, and because it’s based on AI, it’s getting better the more it is used,” said Steve Spear, Visual Data’s Senior Vice President of Operations. “We are starting to revolutionize the restoration process of library and archived content. What used to take many weeks now gets done in a few hours—and the process is much more economical. But to make it work, we needed a storage and processing environment to take advantage of our new capability.”

The plan was to buy high-performance storage to support the company’s traditional core business and its new Matchmaker business. The question was, which technology to use.

Visual Data started by looking at every high-performance storage vendor in the market, but quickly narrowed its choice down to three. The final selection was made based on a combination of factors: performance, scalability, security, manageability, technical support and cost-effectiveness.

The Quantum Xcellis platform was chosen based on its flexibility and performance characteristics as a true high-performance hybrid storage solution. The configuration included an Xcellis Workflow Director running StorNext software, QXS RAID arrays, a Fibre Channel fabric and an Xcellis Workflow Extender to provide access to the files over the company LAN.

In performance testing, the Quantum system did everything that Visual Data needed and more. The team was able to run 4K scanning, UHD transcoding and a 4K playback all simultaneously, and everything worked perfectly without a single dropped frame. StorNext’s security features are also important to Visual Data, delivering a higher level of granularity that enables the company to provide studio clients an improved audit trail.

Quantum’s manageability allows the team to increase capacity by six times without adding staff; and the Xcellis scalability lets Visual Data add performance and storage capacity independently— “which means we’ll be able to grow to meet future demands while avoiding forklift upgrades,” Spear noted.

On overall economics, the Quantum solution was also the clear leader. “Quantum’s cost-per-TB beat everybody,” Spear said, “and the system’s flexibility will help us keep costs low going forward.”

Finally, the Quantum system gives Visual Data the option of using Fibre Channel for some clients, while using Quantum’s Distributed LAN Client (DLC) for others. DLC combines Ethernet connectivity with block-based transfer that outperforms NFS and SMB to boost performance and keep costs low. And Quantum’s approach to the cloud aligns nicely with Visual Data’s plans to leverage cloud storage in the future.

“We like the fact that Quantum gives us the most choices,” Spear says. “Quantum supports third-party commercial cloud offerings, and the StorNext file system lets users see all the content, no matter where it is, through one interface—a really great feature.”

ProductionHUB ProductionHUB Logo

Related Blog Posts


There are no comments on this blog post.

You must be logged in to leave a comment.