Dynamic server-side content stitching for personalized streams.
summary
Claims a production-ready, stateless pull-oriented architecture that stitches playlists into player-agnostic streams, uses a tiny dereference MP4 cached manifest, and applies penalty-based ABR track matching to pick the least-disruptive tracks. Timeline gaps get fixed by adjusting video frame timings and padding audio with silence, and mismatched codecs/aspect ratios trigger just-in-time transcoding where a payload of two MP4s (desired vs actual) is hashed for cacheable transcoding and returned with cropping/letterboxing and AVC3 sample-entry signaling.
AI-written summary · about
chapters (7)
related talks
25mA Stateless Service for Audio Processing
16mCDN Challenges of HTTP-based Low Latency Live Streaming Delivery
22mDesigning for Live Stream Failure with Seamless Switching
21mThe challenges of deploying Apple's Low Latency HLS In Real Life
12mImplementing HLS/DASH Content Steering at Scale
18mHow we accelerated Live 2 VOD delivery from 15 minutes to under a minute
21mPlayback on Edge - Or how I learned to stop worrying and love serverless.
10mMoving Live Video Quality Control from the Broadcast Facility to the Living Room