Adventures with VP9
summary
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 popular videos get transcoded, rate control switches from CRF to target-bitrate and filters out a few percent of worse encodes, and a device capability framework aggregates model+OS data to pick system decoder or in‑app path. GPU-based color space conversion halves decode time (40ms → 20ms) and raises successful playback coverage.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:11WHY VP9? BETTER CODINC EFFICIENCY
- 1:51VP9 BITRATE CONTROL
- 3:22PRODUCE VP9 ENCODINCS AT SCALE
- 4:13INITIAL RESULT ON INSTAGRAM ANDROID
- 4:59THE FRAGMENTATION PROBLEM
- 5:30SYSTEM DECODERS
- 6:05IN-APP SOFTWARE DECODER
- 6:37DECODER CHOICES
- 7:00DEVICE CAPABILITY FRAMEWORK
- 8:10VPX DECODING TIME
- 9:10FASTER SOFTWARE DECODER
related talks
26mNew video coding techniques under consideration for VP10 - the successor to VP9
29mHow to build a high quality 'live' HD VP9 Encoder
26mDASHing Into the Wild with VP9
11mFrom Artifacts to Art: A Data-Driven Guide to Android Video Encoding
31mAV1 Update
18mOn-device Video Playback Upsampling
33mNext-gen video codecs and HDR formats in Android Multimedia
23mDecoder Complexity Aware AV1 Encoding Optimization
