Archive Efforts Come To Fruition
Achieving ROI from old content should only get easier with the proliferation of new AI automation tools, including agentic AI, to more efficiently tag, index and search for relevant video. But human involvement is still essential throughout the process, from delicately handling aging tapes and film canisters as they are digitized to verifying the provenance of decades-old news footage when fulfilling licensing requests, said archiving experts who gathered last week for the TVNewsCheck webinar “Mining the Archive for Content and Revenue,” moderated by this reporter.
Hearst Television completed in late 2025 a six-year project to digitize its extensive tape library, a big task made more challenging by the COVID-19 pandemic. But now the station group is reaping by the benefits by repurposing vintage footage to create special programming for its broadcast channels around holidays or anniversaries of big events, such as the 1995 Oklahoma City bombing; produce regular streaming fare like the Hometown Tragedy true-crime docuseries, and generate a variety of other content for its websites, mobile apps and social media.
“We’ve had a rich history beyond just presenting the news, of creating special programming,” said Ernie Mourelo, VP of digital news, Hearst Television. “In Boston and New Hampshire, we have Chronicle, which is a lifestyle program. We’ve had lots of other programming like talk shows in different markets, and it’s really helped us unlock a lot of that content from the past.”
At the start the process was “very human-involved” with staffers manually loading tape machines, Mourelo said. But as it progressed Hearst was able to take advantage of some AI tools to process video on the back end and automatically create metadata, including scene detection, face detection and audio transcription.
“It really helped us enrich the metadata for those stations that we digitized once we started working with that company [Moments Lab], and it’s actually helped us kind of scale up and be much faster,” Mourelo said. “So rather than humans going through and kind of reviewing each piece of footage that we digitized to recognize its value, AI has helped us really home in on some of the higher-value content. Obviously, we give the parameters of what we’re looking for, but it’s helped us identify the content that’s going to bring us the most value going forward.”
While it’s early, Mourelo said, Hearst has found that about 30% of the content it has digitized has value for either licensing or producing original programming. Early use cases were simply stations being able to quickly find video from 20 to 30 years back to enrich a story they were producing for their daily newscast “without having to go down to a basement and ask somebody to look for a tape.” Some long-legacy stations now have recurring segments on social media or in their newscasts that feature archival footage.
“We have a pretty regular system to produce content nationally around big trends,” Mourelo added. “Think Pokemon, think holiday toys. And we’re really working across teams, so our consumer unit will tap into the archives, our data team will tap into the archives to accessorize, for lack of a better word, the content they’re producing.”
Fox is nine years into an effort to digitize all the archive content from its owned stations and news networks, including analog and digital tapes in various formats and decades of film. The company has turned to outside vendors like Memnon and Iron Mountain to handle the process, sending off truckloads of tapes at a time, but has a dedicated team in Fox Archives to run its third-party licensing business and provide content to internal clients.
Unlike Hearst, Fox is far from the finish line, said Ben Ramos, VP, Fox archives & emerging technology. But it recently completed digitizing every 1-inch, 2-inch and U-matic tape in its library, representing around 100 years and 100 million minutes of content, which Ramos called a “big milestone for us.”
The next step for Fox is digitizing its extensive film repository, Ramos said, then transferring “millions” of Betacam, DVCAM, DVCPRO and VHS tapes, DVDs and hard drives. The goal is for everything to make its way into the Fox Media Cloud, an enterprise-wide pool of content hosted on the AWS public cloud that eventually will provide access to every Fox business unit. (In addition to the AWS storage, there is also an offline, airgapped version of the Media Cloud hosted on LTO tape.)
“We have a long, long way to go,” Ramos said. “And from an AI perspective, we’re just barely dipping our toe in any world of AI. We’ve been using speech-to-text on every asset that’s been going into Fox Media Cloud, which is our AWS S3 buckets. And that’s actually been a big step in searchability, just the speech-to-text portion of the equation, because usually the anchor or the reporter says something that is significant to the starting search point.”
Ramos said that he expects significant AI progress to come in 2026 for Fox Archives from the Fox Media Cloud’s software architecture team, as they increase the number of AI models that content is run through.
“It will certainly expand the searchability, and then the generative capabilities and the productions will be super-charged as well,” Ramos said. “But still, it’s going to be very human-based, and we’re going to still need to do a lot of verification of whatever’s coming out. But search, at least, is going to get really, really good, in a heartbeat.”
One of the archive’s biggest successes so far are Fox Flashbacks, 30-to-60-second mini-documentaries that highlight big stories of the past. The segments first started at WNYW New York, championed by SVP & GM Lew Leone, and have now been adopted across the 18 Fox-owned stations as well as Fox News Channel, Fox Business, Fox Weather and Fox’s various digital apps.
Ramos’ team at Fox Archives provides Fox Flashbacks segments three different ways: as a fully finished, ready-to-air package complete with Fox branding and Fox sound; as a clean, “white-label” version without branding or music; and then as just the four raw assets associated with the story.
“You just listen to what your clients want and need, and you try to spoon-feed it for them because every newsroom needs to hit air as fast as possible,” Ramos said. “You try to deliver it in the easiest way possible.”
Sinclair has changed the way it thinks about its archive as part of its overall “Content Center Transformation,” in which all of its newsroom tools are moving to the cloud, said Mike Palmer, Sinclair AVP, advanced technology and media management. It has discarded the notion of separate buckets of archive and production content.
“That’s not the way that journalists tend to work right now,” Palmer said. “They don’t care where things are located. They want to do a search; they want to find related material.”
Sinclair already has 25 million assets that were “born digital,” i.e., file based, and those have already been moved to the cloud. Sinclair plans to run those clips through speech-to-text algorithms this year to enrich their metadata, which will make them easier to be found and repurposed by journalists and digital content creation teams producing podcasts, series and documentaries.
To contain costs, Sinclair has taken a different approach than Hearst or Fox with its on-the-shelf tape archive, which represents about the same number of clips (25 million) as its digital-native library (but a smaller number of discreet tapes). Instead of digitizing everything, Sinclair first takes pictures of the tape labels and any related cut sheets with a smartphone. It then indexes the pictures and runs them through AI tools from Moments Lab, including Optical Character Recognition (OCR) and an LLM, to create usable metadata.
The LLMs are trained on “huge amounts of data over long periods of time,” Palmer said, which allows them to find known news stories from relevant dates that correlate to the keywords found on the tape label. From that, they can synthesize a description of the likely story along with a probability and a potential value, identify keywords and suggest categories of possible content that Sinclair might want to create.
“Doing that, we are able to very quickly go through many more tapes than we could, per unit time, to digitize,” Palmer said.
This metadata is available to the newsroom systems, allowing journalists to see suggested content as they’re working on a story. If a recommended clip hasn’t been digitized already, they can create an order whereby the tape will be digitized and the content run through further AI models.
“Our metric that we’re really looking to optimize is what is the ratio between what is being used and what we have digitized, not what is on the shelf,” Palmer said. “We want to keep track with good metadata of everything that we have on the shelf first and then choose selectively the content that we’re going to digitize. And then from that, we have reuse and distribution.”
While third-party licensing is still important, Sinclair has decided over time that its internal content teams are the archive’s top priority.
“If it’s our content and it’s valuable, it doesn’t make sense to give it to somebody who’s going to compete with us and sell it to them,” Palmer said. “We want to have first crack at using that content and increasing the value of our own products.”
Veritone has been helping broadcasters organize and manage their archives for over a decade. The Irvine, Calif.-based company processes about 150,000 hours of new content each day. It not only supplies technical infrastructure with its AI software but can also provide full-service representation for broadcasters who want to license their software to third parties but don’t have the sales representatives to support it.
“We can orchestrate 900 different models to create the right metadata at the right price,” said Paul Cramer, Veritone VP of sales, media, entertainment and sports. “We might look at an archive and say, ‘OK, for the portion of the archive that’s sports content, we’re going to use a high motion video model that’s better at detecting things inside of sports, but that’s cost prohibitive to run on news. When it’s news content, let’s use a different model.’”
All that metadata is put into a data lake and is API-accessible, which futureproofs customers’ investment in metadata by making it available to other applications outside of Veritone’s software platform. Customers don’t need to re-scan their content when they’re working with another vendor or another use case in the future, Cramer said.
Veritone’s software layer can also service both internal and external clients. Internal clients could social media teams creating YouTube shorts or social media posts, sales teams that need to recap advertiser integrations or journalists looking for footage they can reuse as B-roll in a segment they’re producing or searching for an old interview for fact-checking purposes.
“We want to democratize access to all of that content in a role- and permission-based foundation,” Cramer said. “On the external side, we have the option to deploy a storefront. We have clients that will literally put a portion of their content publicly available. It’s all watermarked and locked down. But now people who want to license clips, want to find content, to see if it’s available, can find the content they want and put it into a shopping cart. It then sends it to the licensing team at these broadcasters to be able to decide if they’re going to negotiate that and license it.
“We go all the way from that self-serve model to a completely managed service model, where we can actually go out and represent the content in front of documentarians and filmmakers and commercial production companies, handle the rights and clearances, the takedown requests and do everything in between,” he continued. “We work at a full spectrum from self-serve to managed services.”
The post Archive Efforts Come To Fruition appeared first on TV News Check.
It's not every day that you need to shop for formal wear, but when it's…
A newly identified malware campaign is raising serious concerns across the cybersecurity community by delivering…
A newly identified malware campaign is raising serious concerns across the cybersecurity community by delivering…
Security researchers have uncovered a highly sophisticated attack campaign that weaponizes a legitimate, digitally signed…
Security researchers have uncovered a highly sophisticated attack campaign that weaponizes a legitimate, digitally signed…
FORT WAYNE, Ind. (WOWO) — Indiana will see a short-lived stretch of improving and warmer…
This website uses cookies.