Document Processing · Enterprise Knowledge Search
Imperial War Museums needed to make more than 20,000 hours of oral history recordings accessible despite challenges such as regional accents, military terminology, multiple languages, and inconsistent audio quality.
Reported impact
IWM made a large oral history collection more searchable and accessible for the public and researchers, opened new commercial licensing possibilities, and proved a scalable archival workflow with high transcription and speaker-diarization accuracy.
Imperial War Museums, Capgemini, and Google Cloud built a generative AI pipeline to turn a hard-to-search oral history archive into a searchable, cited research experience. Capgemini created IWM’s Google Cloud AI environment and developed a workflow that transcribes and translates audio recordings, extracts metadata, and uses Gemini 2.0 to generate interview summaries plus key people, places, and events. The application lets users search interviews through text transcripts and metadata, listen while reading synchronized transcripts, and use an “Ask a question” interface to submit natural-language queries against interviews. Responses are grounded in the source material and show citations so researchers can trust where answers came from. Prompt engineering and model tuning helped the team handle accents, multiple languages, military terminology, and inconsistent recording quality while scaling a job the museum estimated would otherwise have taken more than 20 years manually.
Audio recordings are transcribed and translated in a Google Cloud pipeline, metadata is extracted and passed into Gemini 2.0 for summaries and key-entity analysis, and the resulting transcripts and metadata power searchable interfaces plus the Ask a question experience with cited answers.
"Capgemini developed a pipeline to process this wealth of audio files – extracting metadata and passing it through Google Gemini to generate comprehensive summaries of interviews"
"Imperial War Museums (IWM) needed to transcribe and make 20,000+ hours of historic oral history interviews held on audio files more accessible."
"The innovative “Ask a question” functionality allows users to ask natural language questions about any interview and receive answers drawn directly from the content."
"With a remarkable 99%-word accuracy and 94%-speaker diarization ... the solution represents a scalable approach"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Capgemini Sample Dec 17, 2025 Amsterdam City Archives Amsterdam City Archives needed to make a massive paper archive usable for the public despite handwritten documents, outdated Dutch, and limited search tools that made discovery difficult. 50 kilometers of paper archives AI Search · Chat with History · Handwritten Character Recognition (HCR) 
OpenAI Sample Aug 31, 2026 Polimill Polimill needed to reduce administrative burden across fragmented Japanese municipal systems so public employees could find prior decisions, prepare assembly responses, and deliver services without each municipality… 3-5x development speed increase from Codex and OpenAI hands-on support Codex · GPT models · OpenAI API 
NVIDIA Sample Jul 7, 2026 Public Ministry of Rio Grande do Sul (MPRS) MPRS needed to manage rising legal caseloads, archived police inquiries, and public-service requests more efficiently, but the volume of legal documents and the need for compliant, locally controlled analysis made… 497 municipalities Archiving Intelligence Assistant (AIA) · Citizen Access Agent · Justice Intelligence platform 