Video @Scale Fall 2022
who it's for
streaming engineers, media infrastructure engineers, video engineers, ar/vr engineers, platform engineers, network engineers, ml engineers, mobile video engineers
topics
real-time AR on constrained devices; distributed AR engine MetaSpark; gpu-backed edge inference for AR; WebRTC and QUIC transport; live media over QUIC; low latency ingest techniques; redundancy-first Rush ingest; dynamic CDN switching with DASH; scaling AV1 end-to-end delivery; hdr tonemap on Android; hdr workflow on iOS; gaze-driven foveated streaming
| title | author | length |
|---|---|---|
| Welcome Remarks | Abhinav Kapoor, Venus Montes | 05m |
Conference opens with a roadmap of concrete advances: platform teams show automated pipelines that remove backgrounds without green screen and suppress audio… | ||
| Building Real Time AR experiences at Scale on Constrained Devices | Pranav Saxena | 32m |
Claim: run GPU-backed ML inference and 3D rendering on nearby edge servers to hit sub-200 ms SLA for shared AR calls and avatar rendering. Uses MetaSpark as a… | ||
| Bringing Interactivity to Videos | Yurong Jiang | 14m |
Built a timed metadata pipeline that carries interactive objects from creation through delivery and playback, with live events packed into emsg MP4 segments… | ||
| Democratizing AR: Building an AR Platform for Everyone | Ben Hazan | 15m |
Arti.AR delivers a cloud-based AR-for-video platform that streams Unreal Engine renders to browsers using Unreal Engine in the cloud, WebRTC and STUN/TURN… | ||
| Video @Scale 2022 LIVE Q&A #1 | Venus Montes | 18m |
Edge compute hosts GPU backends to offload heavy AR ML (GANs, avatar generation) and reduce device work, using geographically distributed edge servers plus… | ||
| Live Media Over QUIC | Luke Curley | 19m |
Twitch builds Warp and runs it in production for distribution while Meta uses RUSH for ingest, both built on QUIC to replace TCP's head‑of‑line limits. The… | ||
| Lessons Learned: Low Latency Ingest | Adam Roach | 17m |
Caffeine reimplements its ingest gateway in Go using Pion, libav, HighVisions go‑SRT, Cisco open hq64 and Comcast go‑TS and ships it in August. A decade‑old… | ||
| Delivering Reliable Live Streaming Over Unreliable Backbone Networks | Jordi Cenzano, Thomas Higdon | 18m |
Meta implements a redundancy-first ingest path that translates RTMPs to Rush in POPs, writes video as configurable HTTP chunks to distributed storage OIL, and… | ||
| Video @Scale Fall 2022 LIVE Q&A #2 | Abhinav Kapoor | 23m |
Live video shifts toward QUIC-based delivery with experiments reporting warp in production and Meta using Rush for ingest while pushing an IETF draft and… | ||
| Gaze-Driven Video Delivery: Science Fiction or Viable Link to the Metaverse? | Yuriy Reznik | 18m |
Gaze-driven delivery claims up to 500x pixel reduction for displays and even up to 8000x reduction for VR by encoding a 1° foveal patch instead of full frames… | ||
| Cloud Streaming in Metaverse | Naizhi Li | 19m |
Cloud streaming runs the full app or selected render objects on server VMs (example: Crayta, Horizon Worlds) and returns encoded video to clients like browsers… | ||
| Building a Professional Video Editor on the Cloud, Powered by Machine Learning | Anastasis Germanidis | 20m |
Runway converts every ML output into HLS model streams so the browser can request processed segments on demand; interactive tools use a click-to-keyframe model… | ||
| Video @Scale Fall 2022 LIVE Q&A #3 | Venus Montes | 18m |
Cloud-driven metaverse delivery claims 25 Mbps minimum for full offload and up to 5–6× bandwidth savings using gaze-driven rendering, but end-to-end latency… | ||
| Approach to HDR and Tonemap on Android | Bhumi Sabarwal | 12m |
Meta engineers implement a frame-level shader tonemap that runs on decoded YUV frames to convert device-captured HDR (PQ or HLG, BT.2020, 10–12 bit) into SDR… | ||
| HDR at Instagram: The iOS Story | Chris Ellsworth | 10m |
Instagram adds full HDR upload and playback support for iOS, moving from H.264 to HEVC main 10 profile for 10‑bit capture and preserving HDR for IGTV/Reels… | ||
| Scaling AV1 End-to-End Delivery at Meta | Ryan Lei | 18m |
Meta describes an end-to-end AV1 rollout for Reels using SVT-AV1 with 13 presets and a two-stage hybrid hardware+software pipeline: first-stage fast HW/AVC… | ||
| Content Steering with MPEG DASH | Will Law | 26m |
A DASH Industry Forum spec proposes a JSON steering server referenced in the MPD that uses baseURL serviceLocation labels and a contentSteering element with… | ||
| Video @Scale Fall 2022 LIVE Q&A #4 | Abhinav Kapoor | 22m |
Real-time GPU shader tone mapping runs for previews and uploads with latency comparable to SDR uploads, using an ITU-R tone mapping recommendation to preserve… | ||
| nothing matches | ||