Highly-Efficient SVT-AV1-based Solutions for VOD Applications
summary
SVT-AV1 adds CRF and multi-pass VBR modes with a three-pass VBR (fast IPP, CRF-like middle, final) and TPL-based QP modulation, evaluated on the elfuente Netflix set (8 min, ~140 shots). Shot-based convex-hull encoding uses vmaf/psnr/ssim via vmap and a convex-hull tool; M0–M12 presets span very-slow to very-fast. Results show VBR within a few percent of CRF, M12 yields ~25% lower bitrate vs x264 and is ~14× faster than x264 very slow, while fast-preset convex-hull re-encoding gives ~1.23× speedups (example M11→M6 ≈3×).
AI-written summary · about
chapters (9)
- 0:00SVT-AV1 for VOD intro
- 1:11SVT-AV1 encoder updates
- 3:40CRF mode improvements
- 8:40VBR multi-pass algorithm
- 11:12Encoder performance eval
- 14:47Convex hull encoding
- 16:31SVT-AV1 presets & savings
- 17:34Acknowledgments & configs
- 18:38Q&A
related talks
18mScaling AV1 End-to-End Delivery at Meta
20mNotes on AV1 productization and royalty-free video codecs outlook
31mAV1 Update
17mDelivering better video encodes for legacy devices
20mStreaming at 250 kbps: raise your expectations
25mAV1 for RTC: Current and Future
29mHow to build a high quality 'live' HD VP9 Encoder
23mAV1 Encoder for RTC