Imperial War Museums (IWM), Capgemini, and Google Cloud have today announced a landmark collaboration on the successful AI-powered transcription and translation of more than 20,000 hours from IWM’s oral history collection. This transformative project will make invaluable firsthand accounts of 20th-century conflict easily accessible to the public, researchers, and educators globally, using advanced generative AI to transcribe, translate and enable interactive exploration of the archives.
This is the first time a UK museum has used AI transcription technology at this scale to create a resource available to users to search, explore, and use sound recordings. IWM plans to release this new technology to the public via its website later this year. A new platform will complement the recordings and resources already available, where users can search records for more than two million collection items.
A significant portion of IWM’s oral histories from 1945 to 2000s—comprising roughly 8,000 interviews with service men and women—were previously only available as audio files. This made accessing them a time-consuming process. These recordings, capturing unique experiences of conflict, presented additional challenges including a wide range of expressions from the time they were recorded, specialised military terminology, and varying audio quality.
Capgemini, working with Google Cloud, developed an innovative solution that makes these recordings part of IWM’s wider oral history collections more accessible. The project used a sophisticated pipeline built on Google Cloud, employing Google’s Gemini models for transcription and analysis. Beyond basic transcription, the system extracts metadata such as the names of people, places, and military units and generates comprehensive written summaries of interviews, highlighting key events and themes in the recording. This process, which would have manually taken an estimated 22 years, will instead take only a few weeks and will significantly enhance how users can access and search this extensive oral history collection, quickly finding the material they are looking for and moving easily between written transcripts and the original audio recordings.
“This project is a big step forward in our mission to broadening access to our vast collections,” said Nick Hodder, Director, Digital Engagement and Transformation at Imperial War Museums. “Our expert curators have been fully involved in this work, ensuring the technology delivers very high levels of accuracy, including understanding and interpreting accents, historical facts and military terminology. This landmark collaboration between IWM, Capgemini and Google Cloud is a significant innovation and a first for a UK museum.”
The technology boasts impressive results, with 99 per cent word accuracy and 94 per cent speaker diarisation (partitioning audio according to the identity of the speaker) on transcription tests. The application allows users to search across interviews using free text, listen to recordings with synchronised transcripts and access AI-generated written summaries through an easy-to-use interface. A groundbreaking “ask a question” function enables users to ask natural language questions about any interview, receiving answers drawn directly from the content, with citations, ensuring accuracy and facilitating a wide range of research needs.
“We are incredibly proud to partner with Imperial War Museums and Google Cloud on this culturally significant initiative,” stated Steven Webb, UK Chief Technology and Innovation Officer at Capgemini. “This project showcases the profound impact of generative AI in unlocking historical archives and making them accessible in new and engaging ways. It’s a testament to how technology can connect us more deeply with our past to inform our future.”
John Abel, Managing Director, Office of CTO, Google Cloud added, “Google Cloud is committed to empowering organisations like Imperial War Museums with AI tools that can transform how we interact with history. The use of Gemini models to process and understand such a vast and nuanced audio collection demonstrates the sophisticated capabilities of generative AI to overcome complex challenges and deliver meaningful outcomes.”
IWM plans to expand its AI capabilities in the future, combining AI analysis with human expertise to significantly improve the ability of a broad range of potential users to access and engage with parts of its collection, including researchers, academics and the wider public. The technology also supports museum curators, creating new ways to interpret powerful stories of conflict that were not previously available.
Notes to editors
Case studies of potential uses
IWM’s sound archives are a valuable source of information for family historians looking to understand more about a relative’s wartime experience. By searching the new platform, it will be simple to find first-person accounts of people who shared a similar experience to a relative. Available as written transcripts and the original audio recordings, these oral histories will add much greater life and depth to a user’s understanding of their relative’s experience than any official records. The new platform will complement IWM’s existing online resources, including Collections Online, Lives of the First World War and War Memorials Register.
IWM’s audio interviews have long been a rich resource for anyone wanting to find out more about the realities of wartime experience. While these have long been available to listen to as audio files on IWM’s website, there have been no written transcripts so they cannot be accessed by deaf people or others who rely on written material. This valuable historic testimony will now have a significantly greater level of accessibility through searchable, AI-generated text versions.
Researchers looking for examples of oral history testimony to feature in documentaries, podcasts, magazine articles and books will find it much easier to find the information they are looking for. Previously this would have involved listening to many hours of sound recordings from start to finish to discover relevant clips. Now a simple keyword search will take a researcher straight to the relevant audio and text transcript, which will then be available for commercial licensing in both written and audio formats.
Academic researchers looking for first-hand accounts to illustrate and inform books, theses and articles will find it far more straightforward to discover useful and relevant material. By using the AI chat function to ask questions such as How did you feel after visiting Hiroshima? or How did you celebrate the end of the war? within a specific transcription they will quickly be taken to citations for the most relevant sections, making research far quicker and more fruitful than was previously possible. IWM’s sound archive is one of the largest and most respected repositories of primary source material in the world, so increasing access through cutting-edge AI technology will be hugely beneficial to the academic community.
