From Transcription to Live Music: Gemini's Audio Stack
summary
One API call to Gemini 3 Flash preview returns speaker labels by name timestamps emotion tags language detection with English translation and a summary; the demo shows structured JSON output ready for UI. A small set of ~30 base voices is steered by a director’s note to produce targeted accents and pacing. A real time sound to sound multimodal model Gemini 3.1 Flash Live bakes reasoning into audio and calls Lyria 3 live to generate a German techno schlager about the UK startup scene and it works on stage.
AI-written summary · about
chapters (9)
- 0:00DeepMind audio work
- 1:18Recent audio models
- 3:04Gemini 3 audio
- 7:40Speech generation styles
- 12:13Real-time multimodal audio
- 15:43Developer tools
- 16:05Music generation with Lyra 3
- 17:29Live demo: Life Jukebox
- 19:02Closing remarks
related talks
21mVeo 3 for Developers
41mBuild intelligent Android apps with Google's AI
20mWhat's new in Machine Learning Keynote
08mDemo: Gemma on-device with MediaPipe and TensorFlow Lite
22mAI-Powered Creative Experiences
20mUsing Generative AI to Increase the Impact of Your Data Science Work
25mCreative video editing with Jetpack Media3 and Gemini
11mapproach to generative AI