Building an ML-powered Android Live-streaming App
summary
Builds a working Android live-streaming app that uses LiveKit + WebRTC for rooms, tracks and data channels, a dev LiveKit server plus an auth/off-service that mints join tokens, and an Android client using Compose, Retrofit and the provided VideoTrackView. Prototypes push pixel frames into an RTC video source, run TensorFlow Lite face landmarker and OpenCV background removal, limit rendering to ~30 FPS, and demonstrates real projects (Bell race‑car demo, Instacart produce offload) with Python agents that successfully overlay inference on live streams.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:54Consulting to innovation lab
- 8:07WebRTC and LifeKit matter
- 13:39LifeKit server demo setup
- 17:52Android live streaming architecture
- 26:36WebRTC performance basics
- 31:16ML integration with LifeKit
- 36:38ML prototyping use cases
- 43:30Q&A
related talks
44mSeamless mobile real-time communication with WebRTC
26mReal-time Audio Streaming Dreams with ReactNative and Kotlin
08h 56mLivestream | Day 1, theater 1
34mTalk: Developing Media Streaming Android Apps..
29mFrom TensorFlow to ML Kit: power your Android application with ML
44mOff to the Races with Android
47mZero to App Live coding an Android app on Firebase in Kotlin
31mSupercharging Android Apps with Tensorflow