Most oral history collections are barely accessible to the people they belong to.
We're building with cultural workers, creators, and institutions. Get early access or partner on a pilot.
Partner with usMost oral history collections are barely accessible to the people they belong to.
That's not a provocation. It's a description of how the majority of cultural institutions currently operate. Recordings sit in climate controlled storage. Transcripts, where they exist, are keyword searchable at best. A researcher who knows exactly what they're looking for might find it. A community member following a thread of family memory almost certainly won't.
The promise of AI in the heritage sector isn't about replacing archivists or automating cataloguing (though it can help with both). It's about closing the gap between what an institution holds and what its communities can actually reach. An oral history AI platform, built with the right principles, turns a static collection into something people can genuinely converse with.
The British Library holds over 30,000 oral history recordings. University special collections across the UK house thousands more. Migration stories, industrial heritage, dialect recordings, community testimonies gathered over decades. The Black Cultural Archives in Brixton safeguard irreplaceable first-person accounts of Black British life.
These collections represent extraordinary care. Fieldworkers, archivists, and community volunteers spent years building them.
Preservation and access aren't the same thing. A recording preserved on magnetic tape in a temperature controlled vault is safe. It's also, for most practical purposes, invisible.
Traditional oral history preservation follows a linear path: record, transcribe, catalogue, store. Each step is labour intensive and expensive. Transcription alone can take four to six hours per hour of audio. Cataloguing requires specialist knowledge. The result is a bottleneck where material enters the archive far faster than it can be made discoverable.
This isn't a failure of effort. It's a structural limitation of the tools available.
Artificial intelligence introduces three capabilities that matter for oral history preservation.
Transcription at scale. Modern speech-to-text models handle accents, dialects, and multilingual recordings with increasing accuracy. What once required weeks of manual work can now produce a working transcript in hours, with human review focused on correction rather than creation.
Semantic search. Traditional keyword search requires you to know the exact terms used. Semantic search lets someone ask "stories about arriving in London in the 1960s" and surface relevant material even when those precise words never appear in a transcript.
Conversational retrieval. RAG technology goes further . Users ask questions and receive grounded, cited responses drawn directly from the archive. Not hallucinated summaries. Actual excerpts, with provenance intact, in response to natural-language questions.
Together, these capabilities form what we call a Living Archive. A collection that's searchable, citable, conversational, and community-controlled.
From the ecosystem · Studio Note
How RAG technology makes cultural archives come alive
Your grandmother's voice is recorded. Your community's textile patterns are photographed. All of it exists. None of it is findable.
Threads of Memory, our pilot project, is building this with a collection of oral histories from migrant makers working along London's Weaver Line. A conductive textile map triggers oral histories when touched. A conversational interface lets anyone ask questions of the archive and get cited responses drawn from real testimony. The installation and an audio walk along London's Weaver Line are in development now.
Oral histories aren't training data. They're testimony, offered in trust.
Crucially, the project operates under community governance. Access permissions, content moderation, and data handling are all determined by the contributing communities, not by the institution hosting the technology.
From the ecosystem · Current Project
Threads of Memory: Haringey, London
Touch a textile, hear a story. An immersive archive rooted in London's Weaver Line, keeping the voices of textile workers, migrants, and makers alive.
The generation that recorded many of the UK's most significant oral history collections is ageing. Their children and grandchildren want to access those stories, not through a reading room appointment, but conversationally, on their own terms, when the question arises naturally.
There's a window here. The people who can verify, contextualise, and enrich these recordings are still with us. The technology to make that enrichment permanent is now available.
Waiting is a choice with consequences.
For institutions considering this path, the starting point isn't the technology. It's the relationship.
Start with community consultation. Before selecting a platform, transcribing a single recording, or writing a funding bid, talk to the communities whose histories you hold. What do they want access to look like? What are their concerns about AI? What governance structures need to be in place before any data moves?
Audit your existing collection. What's already transcribed? What's digitised but not transcribed? What's still on analogue media? Understanding the state of your holdings determines the technical pathway.
Choose infrastructure that respects your principles. An oral history AI platform should support granular access controls, clear provenance tracking, and community governance from the architecture level, not as an afterthought bolted on after deployment.
Plan for sustainability. A Living Archive isn't a one-off project. It requires ongoing maintenance, community engagement, and technical support. Build this into your funding model from the start.
From the ecosystem · Studio Note
What Is a Living Archive? Meaning, Examples & Why It Matters
Most archives are beautifully organised cemeteries. A Living Archive is something else entirely.
If your institution holds oral histories that deserve to be heard, not just stored, we'd welcome the conversation.