Demuxed 2019
2019-10-23..2019-10-24
San Francisco, United States
Demuxed ·
[Demuxed 2015, Demuxed 2016, Demuxed 2017, Demuxed 2018, Demuxed 2019, Demuxed 2020, Demuxed 2021, Demuxed 2023, Demuxed 2024, Demuxed 2025]
| title | author | length |
|---|---|---|
| Is now the time to solve the deepfake threat? | Roderick Hodgson | 27m |
Claim: an explosion of deepfakes is imminent and must be addressed now. Evidence: a convincing demo assembled from Keanu Reeves footage using Facebook open… | ||
| Super Resolution: The scaler of tomorrow, here today! | Nick Chadwick | 24m |
Claims that ML super-resolution beats bicubic by measurable VMAF margins like +5 to +10 VMAF points, but at the cost of heavy compute and a GPU requirement… | ||
| The do's and don'ts about Streaming security... | Javier Brines Garcia | 08m |
Claims daily monitoring of 7000 stations and a catalogue of which protections actually stop signal capture versus which fail; demonstrates how to extract URL… | ||
| Modeling the conceptual structure of FFmpeg in JavaScript | Ryan B Harvey | 19m |
Claims a concrete JS library called pizzonia that models FFmpeg’s CLI as five objects (command, input, output, filter chain, filter node) and uses the same… | ||
| Objectionable Uses of Objective Quality Metrics | Richard Fliam | 26m |
Argues that objective metrics like VMAF PSNR and SSIM can be gamed and gives concrete attacks: tune psnr and similar encoder tweaks that improve scores while… | ||
| RTMP: web video innovation or Web 1.0 hack… how did we get to now? | Sarah Allen | 22m |
RTMP originates from a 1999 prototype and is released in July 2002, designed for Flash-heavy browsers when over 98% of browsers had Flash installed and ~248… | ||
| Large-Scale Media Archive Migration to the Cloud | Konstantin Wilms | 30m |
A small team spends ~1.5 years moving 40 petabytes of studio assets into a cloud-native platform built on AWS, using Terraform/Nomad/Go workers and plain AWS… | ||
| HEVC Upload Experiments | Chris Ellsworth | 10m |
Claimed up to 2x better compression for HEVC versus H.264 proves false at Instagram story bitrates: manual tests plus millions of production uploads show HEVC… | ||
| Scalable Per-User Ad Insertion in Live OTT | Zachary Cava | 25m |
A new paradigm called Server-Guided Ad Insertion decouples ad decisions from cached content by pushing state to clients and using MPD patch updates with DASH… | ||
| SMPTE ST 2110 - Highest Quality, Lowest Latency, Pro Video over IP | Andy Warman | 22m |
ST 2110 describes uncompressed, low‑latency professional video over IP that broadcasters like FOX BBC CBC and NBC already deploy, supports up to 8K and HDR… | ||
| Building an automated testing suite... | Charles Sonigo, Jean-Baptiste Louazel | 28m |
A DIY automated suite that claims to cover 100,000+ combinations actually builds a grammar, an API to enumerate player/media/browser/OS variables, an… | ||
| Edge Compute | Kyle Boutette | 10m |
Edge compute moves request-time JavaScript onto the CDN to run logic before or around cache lookups, claiming reduced origin load and low latency. Examples… | ||
| Designing for Live Stream Failure with Seamless Switching | John Saxton, Shawn Przybilla | 22m |
Describes an implemented encoder synchronization system that makes HLS segments interchangeable so player failover has no gap, using an output locking… | ||
| A/B User Testing When Your CDN Cache Messes With Your Results | Sebastiaan Van Leuven | 19m |
CDN caching doubles cache misses when client-side A/B routes users to different processed assets, which inflates startup latency and depresses watch time and… | ||
| The “technical” in technical project management | Alexandria Shealy | 20m |
An engineering‑centric project management approach claims concrete wins: 5 months to launch HBO Now, delivering live ad‑supported NFL games on Twitter… | ||
| AV1 at Netflix | Andrey Norkin, Liwei Guo | 24m |
AV1 deployment work centers on an open-source encoder called SVT-AV1 developed with Intel since August 2018, published on GitHub Jan 31 with 63 contributors… | ||
| Reinforcement learning for ABR | Sahil Budhiraja | 19m |
Claims a production ABR replaces heuristics with reinforcement learning trained on real throughput traces and exported from TensorFlow Python into… | ||
| The challenges of deploying Apple's Low Latency HLS In Real Life | Marina Kalkanis | 21m |
Reports using CMAF for ingest and a pipeline with Media Excel transcoders, Unified Streaming origin, nginx cache, Akamai CDN and Apple AVPlayer, implement 320… | ||
| Analyzing Video Metrics like Richard Feynman | Ben Dodson | 14m |
Concrete claims and tools replace platitudes: reviews of QoS metrics like rebuffering startup time watch time, use of Anscombe's quartet to show identical… | ||
| A Stateless Service for Audio Processing | Michael Hill | 25m |
Claims a production stateless pipeline for multi‑participant live audio used at Periscope/Twitter that decodes WebRTC opus→AAC transcode per short request… | ||
| Non-Standard Codecs With Standard WebRTC | Bryan Meissner | 11m |
Claims a working path to stream H.264/AAC over standard WebRTC without native codec support by extracting RTP/raw frames on mobile and sending fragmented MP4… | ||
| Edge Transcoding | Ty Bekiares | 37m |
Multicast HEVC over our satellite network to aircraft and on-board machines transcode 12+ channels to H.264, MPEG‑2 and pass-through HEVC using ffmpeg + vaapi… | ||
| CMAF and DASH-IF Live ingest protocol | Rufael Mekuria | 25m |
A DASH‑IF specification defines two live‑ingest interfaces: CMAF ingest (per‑track fragmented MP4 over HTTP POST, one TCP connection per track, long or short… | ||
| Japanese Captions | Gary Katsevman | 09m |
Explains that Japanese captions commonly sit vertically on the right or left and walks through five concrete features: ruby pronunciation annotations, text… | ||
| Lightning Talks | David I'll soon | 43m |
Collection of short demos and deployments with concrete claims: a replay-as-live tool mo replay that handles token auth SCTE external audio DVR looping and… | ||
| QoE Impact from Router Buffer sizing and Active Queue Management | Te-Yuan Huang | 22m |
Experiments compare different backbone router buffer sizes and home-router Active Queue Management (FQ‑CoDel) and report concrete effects: tuning buffer size… | ||
| A paradigm shift in codec standards: MPEG-5 Part 2 LCEVC, higher quality with any... | Guido Meardi | 30m |
LCEVC layers a downsampled base codec plus two residual layers and simplified upsampling to boost any codec’s output while remaining codec‑agnostic and… | ||
| JavaScript? In *my* native video app's core streaming library?... | Steve Robertson | 23m |
Claim: put a tiny Media Source core into native apps and run a JavaScript protocol engine above it that can be versioned with server releases. They build… | ||
| Doing Server-Side Ad Insertion on Live Sports for 25.3M Concurrent Users | Ashutosh Agrawal | 21m |
What’s really the biggest streaming event ever? Superbowl? What could be bigger? Felix’s jump from the stratosphere? Fifa World Cup of 2018?
Let me introduce… | ||
| LiTr (LinkedIn Transformer) - and open source library to transcode and modify... | Izzat Bahadirov | 10m |
LiTr (LinkedIn Transformer) claims on-device transcoding that cuts upload time by lowering resolution and bitrate while preserving viewing quality, and it uses… | ||
| There and back again: reinventing UDP streaming with QUIC | Lucas Pardue | 30m |
QUIC is now an IETF transport being finalized alongside HTTP/3, with 16+ implementations and live HTTP/3 deployments at Cloudflare; the talk claims lower… | ||
| Adventures with VP9 | Haixia Shi, Zen Xu | 10m |
Claim is VP9 yields about 20% BD-rate saving on Instagram UGC but server and Android fragmentation bite: software VP9 costs about 6x CPU versus H.264 so only… | ||
| Three Roads to Jerusalem | Will Law | 37m |
Compares DASH ULL, LHLS and ALHLS and claims a single-set workflow can serve all three; prototypes a pipeline using Big Buck Bunny and ffmpeg nodejs to emit… | ||
| Demuxed 2019 Recap | Josh Brener | 14m |
An actor recaps Demuxed 2019 with jokes and slide mishaps rather than technical claims, demos, or measurements, so expect no technical claims and no… | ||
| nothing matches | ||