AI Powered Transcription Pipeline for Bodhicharya
> 4,000
hours of content transcribed
>1 ,000
sychronized subtitles generated
300
Custom vocabulary terms
Overview
An AI-powered transcription project, producing transcripts and subtitle files for audio and video recordings via a scalable, serverless pipeline hosted on AWS.
We were highly satisfied with the meticulous and insightful work conducted by Lambert Labs
Paul O’Connor, Director, The Ringu Tulku Archive
Opportunity / Customer Challenge
Bodhicharya engaged Lambert Labs to modernize their offering by transcribing their full archive of teachings, totalling over 250,000 minutes across audio and video recordings. In addition, they wanted a scalable pipeline built that could handle newly uploaded teachings.
The aims of the transcription project were to enhance accessibility and enable future projects around AI powered search, translation and more.
Solution
Lambert Labs built a scalable pipeline harnessing AWS services including AWS Lambda, Amazon Transcribe and AWS Elemental MediaConvert. Teachings were transcribed into formatted .txt files for audio, and synched subtitles for video. We made use of the custom vocabulary features of Transcribe to handle specialized, domain specific terminology, and a separate pipeline allows the customer to make ad hoc edits to individual transcriptions and further refine the custom vocabulary. New teachings uploaded to S3 trigger a serverless process to generate the appropriately formatted transcription.
We were highly satisfied with the meticulous and insightful work conducted by Lambert Labs. Their adaptability and flexibility in their approach were instrumental in achieving such a successful bespoke outcome. Notably, their extensive experience with AWS services resulted in substantial cost savings for the project, which we were unaware of. (Paul O’Connor, Director, The Ringu Tulku Archive)
Outcome
The full archive of past teachings, totalling 250,000+ minutes were processed correctly. The pipeline handles all new teachings which are continually uploaded, and the customer is able to continually refine the vocabulary and reprocess transcription jobs.
This was a really rewarding project for both ourselves and Bodhicharya. We got to harness the power of AI on AWS with Amazon Transcribe, and build out a really clean, scalable serverless architecture that suits the customer’s needs perfectly. It’s also fantastic that the transcriptions serve to widen accessibility to resources that Bodhicharya’s audience find hugely useful. (George Lambert, Founder & CEO, Lambert Labs)
About Bodhicharya
Bodhicharya is an international non-proft delivering Buddhist teachings from master Ringu Tulku Rinpoche through their online platform.