RTC @Scale 2022
as the organisers put it
RTC @Scale is a conference for engineers who develop and manage large-scale real-time communication (RTC) systems. The operation of large-scale RTC systems has always involved complex engineering challenges and has attracted much attention in recent months given the explosive growth of RTC in these unprecedented times. The @Scale community focuses on bringing people together to […]
who it's for
real time communication engineers, audio machine learning engineers, video codec engineers, spatial audio engineers, infrastructure and SRE engineers, security and encryption engineers, product engineers for real-time systems
topics
holographic video calling; point cloud and depth compression; VPCC and point-cloud codecs; HRTF spatial audio and server mixing; WebRTC congestion control and QUIC for media; AV1 low-latency encoding for RTC; libaom realtime CBR and layered SVC; autoencoder-based low-rate speech codecs (SoundStream/Lyra); deep noise suppression and DNS-MOS evaluation; speech generative codecs and neural vocoders; end-to-end encryption for large group calls
| title | author | length |
|---|---|---|
| Keynote: Real-time Communication for Today and Future Experiences | Maher Saba | 11m |
Remote Presence is a cross‑app platform built over ~two years to make digital meetups feel like being together, already shipping Watch Together on Instagram… | ||
| Holographic Video Calling | Nitin Garg | 17m |
A practical holographic call pipeline captures RGB+depth, generates point clouds, compresses them and streams over real networks; experiments show VPCC 200×… | ||
| Spatial Communications at Scale in Virtual Environments | Paul Boustead | 20m |
Dolby describes building large-scale spatial voice for virtual worlds, claiming 5,000 players on one server and supporting games with millions of concurrent… | ||
| RTC3 | Justin Uberti | 18m |
A 20-year arc from Skype to WebRTC to post‑COVID real‑time social apps leads to concrete claims and prototypes: HRTF spatial audio to reduce listening fatigue… | ||
| RTC @Scale, Future RTC Experiences Session - Live Q&A | Nitin Garg, Paul Boustead, Justin Uberti, Rahul Gowda | 10m |
Current end-to-end target for audio and video latency is 150 to 250 ms target, though real public networks and holographic processing run higher today; spatial… | ||
| Developing Machine Learning Based Speech Enhancement Models for Teams and Skype | Ross Cutler | 21m |
Microsoft replaces the traditional DSP noise suppressor with a deep learning model that performs 7X better than the previous solution, trained on synthetic… | ||
| Can AI Disrupt Speech Compression? | Jan Skoglund | 21m |
Neural models compress speech down to a few kilobits with much better quality than old low‑rate vocoders and competitive with standard codecs at far lower… | ||
| RTC @Scale, Audio ML Session | Ross Cutler, Jan Skoglund | 05m |
Live Q&A describes deployed deep noise suppression used in Skype/Teams, running on client endpoints with measured CPU cost of 1–2% CPU on Intel and a DSP… | ||
| RTC in the Metaverse @Scale Live Panel | Sriram Srinivasan, Mike Arcuri, Paul Boustead, Cullen Jennings | 41m |
Panel summarizes working demos and concrete challenges: WebEx hologram streams photorealistic 3D people using a light field approach versus point clouds and… | ||
| AV1 for RTC: Current and Future | Zoe Liu | 25m |
Aurora1 AV1 encoder claims real deployments: 1080p zero-latency 15fps on one CPU core and 720p30 real-time on mobile (example iPhone12), with encoder binary… | ||
| AV1 Encoder for RTC | Marco Paniconi | 23m |
libaom AV1 delivers low-latency RTC by using one-pass real-time CBR with no lookahead, a speed 5–10 quality/CPU knob and tile/row multithreading that gives… | ||
| RTC @Scale, Video Session | Marco Paniconi, Zoe Liu | 06m |
AV1 runs in Duo at low resolutions and shows a single-digit battery increase from the non‑AV1 baseline when first deployed about 1–1.5 years ago and then… | ||
| Making Meta RTC Audio More Resilient | Andy Yang | 21m |
Meta RTC uses NACK, in-band FEC, and duplication over WebRTC and measures impact via neteq’s delay histogram and P95 latency. In experiments duplication alone… | ||
| Private Calling at WhatsApp | Xi Deng | 19m |
WhatsApp builds private calling for 2 billion users handling 15 billion minutes a day using a client-centric architecture that keeps call state on clients and… | ||
| Group Call End-to-End Encryption and the Challenges of Encrypting Large Calls | Abo-Talib Mahfoodh | 17m |
Group calls use an SFU so media is routed server-side but is encrypted end-to-end with SFrame and per-sender symmetric sender keys (AES) negotiated via public… | ||
| RTC @Scale, Resilience and Encryption Session | Live Q&A | Andy Yang, Xi Deng, Abo-Talib Mahfoodh | 11m |
Audience Q&A covers concrete tradeoffs: out‑of‑band FEC adds latency so duplication is used for audio while RS code looks promising for video error recovery… | ||
| nothing matches | ||