Why It's Important
The transcription of oral histories is a fundamental part of preserving the invaluable knowledge of Elders for future generations. However, manual transcription is an extremely slow and labour-intensive process. AI-assisted transcription can dramatically accelerate this work, allowing communities to process their backlogs of audio and video recordings much more efficiently. This creates searchable, accessible text archives that can be used to develop educational materials, support land claims research, and provide content for language revitalization. As noted by Library and Archives Canada, making this content searchable is key to its use. By making the wisdom of Elders more accessible, AI-assisted transcription strengthens intergenerational knowledge transfer and supports the cultural and governance needs of the community, which has direct economic benefits in sectors like education and heritage tourism.
History
The practice of recording oral histories began with early wax cylinder and wire recorders. For most of the 20th century, these recordings were transcribed manually by linguists and researchers, a process where one hour of audio could take 8-10 hours to transcribe. The first "automatic" transcription software appeared in the 1990s but was often inaccurate and required extensive training for a single voice. The modern era of AI-assisted transcription is powered by deep learning models developed in the last decade. Services like Otter.ai and advanced AI models from companies like OpenAI (Whisper) have made highly accurate, near-instantaneous transcription accessible to anyone, representing a massive leap forward in efficiency.
Examples
The National Centre for Truth and Reconciliation (NCTR) at the University of Manitoba holds thousands of hours of oral histories from residential school Survivors. They are exploring the use of AI tools to help make these vast archives more searchable and accessible to Survivors, their families, and researchers.
The Canadian Museum of History has an extensive collection of oral histories from across the country. Like many cultural institutions, they are investigating how AI can be ethically applied to the transcription and management of these digital collections.
Many university-based oral history projects across Canada, such as those at the University of British Columbia's Oral History Lab, may use AI transcription services to create first-draft transcripts, which are then corrected and verified by researchers and community members.
Software and Tools
Otter.ai: A popular, user-friendly, web-based transcription service. It provides a generous free tier for individuals (e.g., 300 minutes per month) and can transcribe audio in real-time or from uploaded files. It also has features for identifying different speakers.
Whisper by OpenAI: A highly accurate, open-source AI model for speech-to-text. While it requires more technical skill to set up and run on a local computer, it is free to use and does not require uploading your data to a third-party service, which is a major advantage for data sovereignty.
Audacity: Free, open-source audio editing software. Before transcribing, it's essential to have the best possible audio quality. Audacity is an invaluable tool for cleaning up background noise or amplifying quiet recordings to improve transcription accuracy.
Mukurtu CMS: While not a transcription tool itself, Mukurtu is the leading platform for managing the final transcripts and audio files according to community cultural protocols. You can link a transcript directly to an audio file and set specific access rules for who can view or use it.
AI Considerations
AI transcription is a powerful tool, but it must be used as an "assistant," not a replacement for human oversight, especially when dealing with the sacred words of Elders. The primary benefit is a massive gain in efficiency, which saves on labour costs and volunteer time. The risks, however, are significant. AI models will make errors, particularly with proper names, place names, and specialized cultural terms. If these errors are not corrected, they corrupt the historical record. A "human-in-the-loop" process is essential. Furthermore, using commercial, cloud-based services involves a privacy risk, as you are sending sensitive cultural data to a corporate server. Clearly defining what personal or private information (PII) is contained in the recordings and understanding the vendor's data-handling policies is a critical step.
FAQ
Manual transcription is done entirely by a human listening to the audio and typing it out. AI transcription uses an artificial intelligence model to automatically convert the audio to text. AI is much faster, but less accurate than a skilled human.
It varies. A service like Otter.ai has a free tier and then monthly subscriptions. Using an open-source model like Whisper is free to use, but requires a powerful computer to run efficiently.
The accuracy for Indigenous languages will be lower than for English, especially for languages with fewer online resources. However, models like Whisper can often produce a surprisingly good first draft that is still much faster than starting from scratch.
Yes, many modern transcription tools have a feature that can automatically identify and label different people speaking in a recording.
Before recording, you must get informed consent that covers how the recording and its transcript will be used, stored, and who will have access to it. This is a cornerstone of ethical oral history work.
Pro Tips
Improve your transcription skills by using AI services to generate drafts of oral histories and then carefully reviewing and correcting them. Take the time to add contextual notes and follow cultural protocols regarding access and storage. Working on these projects teaches you about both technology and heritage, ensuring accurate preservation.
Checklist
External Resources
The Oral History Association: An international organization providing best practices and ethical guidelines for oral history projects.
Canadian Oral History Association: A national organization promoting oral history in Canada, with resources and a community of practice.
Digital Archives Toolkit: resource for non-Indigenous archival institutions in Canada working with First Nations, Inuit, and Métis communities and their documentary heritage.
BC Archives: A key provincial resource that holds extensive oral history collections and provides guidance on archival standards.
First Peoples’ Cultural Council: Provides a list of B.C. First Nations Languages