IP excavation: AI breathes life into the archive

Ukrainian tech firm Respeecher has made a name for itself creating synthetic voices for characters in Star Wars franchises and Oscar‑winning films The Brutalist and Emilia Pérez. However, none of its preceding 250 projects were quite as technically and ethically demanding as Endurance. 

The documentary project showcased at the IBC2026 conference session, “AI in Practice: Case Studies Across the Media Value Chain”, required recreating voices recorded on wax cylinders during Ernest Shackleton’s Antarctic expedition. 

“This was a very special technical challenge,” CEO and Co‑Founder Alex Serdiuk said. “Wax cylinders were the only way to record voice then. That’s the data we had.” 

Recovered from the ocean alongside crew diaries, the cylinders formed the raw material for Respeecher’s reconstruction. For Serdiuk, based in Kyiv, the project carried emotional weight. 

 “We know what endurance is from first hand,” he said. “We know what it means to fight a much bigger power.” 

He stressed that Respeecher’s work is built on strict ethics. 

“Quality is essential, but trust is just as important,” he said. “We never let our technology be used by the mass market until we can ensure voices are cleared.” 

He warned of the dangers of unrestricted cloning. “We’ve seen others create fake political content and election intrusions. We never allow that.” 

For Serdiuk, the Shackleton project shows both the power and responsibility of generative AI. 

“Recreating a voice from 100‑year‑old data is extraordinary,” he said. “It’s vital this technology is used in the right way, for the right projects.” 

 AI unlocks 70-years of content 

Likewise, in another IBC session across the halls, ITN and Reuters explored how they are making a 70-year archive fit for the future by preparing it for AI‑driven workflows. 

 “ITN is a news powerhouse,” said Tim Forrest, Head, Content Delivery, Commercial Innovation, ITN. “We look after three major newsrooms and a living archive that grows every single day.” 

Most of the archive is digitised, but the system used to manage it is now a decade old. Attempts to integrate AI into it failed. 

“We needed a platform that served us better today but was ready for tomorrow,” Forrest said. “The challenge wasn’t just internal efficiency. It was making sure we could still sell our archive as technology evolves. We wanted frictionless access delivering profound value.” 

ITN approached Imagen and initiated Project North Star, David Wilkinson, Head of Imagen, Reuters, explained. 

“Self‑build projects are hard to keep moving,” he said. “You have to worry about security, DevOps, scalability, and new technologies.   

Together, the teams began defining ITN’s future archive strategy – Project North Star. 

 “With a huge archive, preservation becomes the default,” Wilkinson said. “Nobody wants to say: ‘We don’t need that.’ But it creates noise. Discoverability becomes incredibly hard.” 

However, both speakers stressed that AI must be applied carefully. 

“AI tools are powerful hammers looking for a nail,” Wilkinson said. “If you haven’t got a nail, it’s the wrong technology.” 

Instead of expensive enrichment models, ITN is starting with mature, cost‑effective tools such as transcription, translation, and scene detection. 

“If you start with the complicated ones, you can burn through a lot of money and learn very little,” Forrest said. “AI puts fear into people. But in a 70‑year archive, there’s comfortably enough work for AI to join in without reducing headcount.” 

Wilkinson agreed. “Archivists who’ve worked these collections for 30 or 40 years know where the nuggets are. AI lets them be more powerful.”