Wasabi adds AI video data tagging to object storage service
Wasabi will soon launch a new service that uses AI and machine learning to search and tag video data. The capability stems from an acquisition the vendor made earlier this year.
The latest object storage offering from Wasabi Technologies adds automated tagging for video data using AI and machine learning capabilities for media and entertainment industries.
Wasabi's AiR service is based on the technology created by Curio AI, which Wasabi acquired in January from GrayMeta Inc. AiR enables customers to store video data in S3-compatible object storage with automatic data tagging and metadata indexing. The tags are deployed by the second and provide facial recognition, speech to text, translation and logo recognition capabilities.
AiR will have no ingress or egress fees, like Wasabi's other storage offerings, and the service won't charge for usage either. Instead, it's offered at an introductory flat rate of $12.99 per terabyte, per month, according to Wasabi spokespeople. AiR is expected to be released within the next several months, with the price likely to change following launch, according to the vendor.
Sifting through and tagging relevant video data is a laborious undertaking, even with automation, and an underserved business, according to Ray Lucchesi, founder and president of Silverton Consulting.
"Video at the moment is ubiquitous, and there's a significant need for [a cataloging service]," Lucchesi said. "Audio is the same way."
Frame-by-frame data playback
Wasabi AiR will target media and entertainment customers first, according to Aaron Edell, senior vice president of AI at Wasabi. Edell was previously president and CEO at GrayMeta.
These customers have begun using cloud object storage as a new form of cloud storage without the physical concerns and constraints associated with tape, he said, but they aren't always taking advantage of the compute capabilities that cloud can bring.
AiR aims to change that by eliminating the need for customers to connect their object storage to external cloud services for tagging capabilities. Customers might use physical tape media to store large, infrequently accessed files such as videos since it's typically cost-effective, Edell said, but object storage can provide a similar savings and connect into more services.
"We don't want customers to think of this as a cold archive that never gets accessed," Edell said.
When uploading video files to the service, the technology will conduct an initial frame-by-frame search and tagging operation. The users can later review this data to refine tagged information and add additional search terms or objects to recognize for later uploads. Customers can also train the AI on their data alone for capabilities such as content moderation.
Wasabi already has sponsorships with media and entertainment companies that could use this capability, such as sports teams including the Boston Red Sox, Edell said.
AiR-ing out old footage
The hype around generative AI has created legal and technical challenges for enterprises, said Steve McDowell, founder and chief analyst at NAND Research. Worse, GenAI has muddied what non-generative AI and machine learning (ML) capabilities are beneficial to enterprises, such as video data tagging automation, he said.
Metadata and tagging of data [are] a critical problem. For that to be in a storage platform, that's a big differentiator for Wasabi.
Steve McDowellFounder and chief analyst, NAND Research
Tagging and metadata generation isn't something many employees would want to take on no matter what business utility might come from the data. Most metadata tagging services work with static documents or images, McDowell added. Some offerings, such as Komprise and its platform, focus on tagging from a data management and classification perspective rather than examining the content itself.
"This is a beautiful use of AI and ML," McDowell said. "Metadata and tagging of data [are] a critical problem. For that to be in a storage platform, that's a big differentiator for Wasabi."
Customers using a hyperscaler such as AWS can piece together similar capabilities, both McDowell and Lucchesi said. But these capabilities might come with associated ingress and egress fees, potentially driving up costs.
Lucchesi said demand will determine if the service can maintain performance for media customers, who typically work with massive video files.
"The question is the types of metadata being created -- are they good enough for what customers want to use?" he said. "Can they sustain it if customers start streaming 8K [resolution] videos?"
Tim McCarthy is a news writer for TechTarget Editorial covering cloud and data storage.
