Artificial intelligence to the rescue of accessibility
summary
Demonstrates using Google Cloud tools to improve accessibility: Video Intelligence extracts scene labels and timestamps, Speech-to-Text transcribes audio in 125 languages with speaker diarization and custom vocab (best-case ~5% WER), Text-to-Speech WaveNet generates natural audio from text (≈220 voices, ~40 languages), and Vision API suggests image labels/alt text. Notes limitations like imperfect transcripts, need for domain-specific vocabularies, and dataset bias issues being tackled by initiatives such as Mozilla Common Voice.
AI-written summary · about
chapters (7)
- 0:00Start
- 10:05SPEECH-TO-TEXT API
- 11:26API DETAILS
- 27:12LIVE TRANSCRIBE
- 28:50LIVE CAPTION
- 33:07COMMON VOICE
- 33:47EUPHONIA PROJECT
related talks
55mBridging the Digital Divide: Empowering Accessibility Through AI Innovations
47mBuilding smarter apps with Machine Learning, from magic to reality
47mAccessibility in the era of generative AI
46mBuilding accessible experiences with AI
30mAI and Accessibility: the Good, the Bad, and the Bollocks
34mAI for Accessibility: Empowering Inclusive Digital Experiences
44mBuilding smarter solutions with no expertise in machine learning
36mCloud AI Services: What they are and how to use them