Using Amazon Transcribe to make content searchable and accessible
summary
Amazon Transcribe plus the Media Insights Engine turns video into searchable, timestamped text and VTT subtitles using S3, Lambda, MediaConvert, Comprehend, Translate and Rekognition; custom vocabularies and IPA fixes improve errors and Amazon A2I adds human-in-the-loop review. NASCAR integrates this pipeline, extracts audio to avoid a 2 GB limit, and reports 99% captions coverage and about 97% cost savings vs vendors. Demo shows editable transcripts, exportable subtitles, and that custom vocab plus CLMs materially raise accuracy. Amazon Transcribe Media Insights Engine
AI-written summary · about
chapters (12)
- 0:00Introduction to AI in Media
- 1:17Challenges of Transcription
- 2:43Amazon Transcribe Overview
- 4:40Key Features and Capabilities
- 7:46Media and Entertainment Use Cases
- 9:57Technical Demo of Services
- 13:35Customizing Transcribe Accuracy
- 17:22Demo Results and Summary
- 18:45NASCAR Success Story
- 21:22NASCAR Workflow and Lessons
- 24:00Future Roadmap and Benefits
- 26:08Conclusion and Resources
related talks
33mMaximize lifetime value of media content
45mAccessible Audio & Video
45mTranscend Language Barriers with Custom Machine Learning APIs: um...
26mEmpower your business use cases
20mAI-Powered Customer Insights for Financial Services
21mBoosting Inclusivity: Closed Captioning & Translations in WebRTC
27mSummarizing & analyzing videos
24mApplications of AI in news and media