VMAF (Video Multimethod Assessment Fusion) is an open-source, machine-learning-based quality metric developed by Netflix to estimate perceived video quality on a 0-100 scale. Unlike older metrics such as PSNR, VMAF correlates well with how humans actually rate quality, which makes it the practical yardstick for encoding decisions.
Its main use is to decide "how much bitrate is enough": per-title and context-aware encoding pick the lowest bitrate that still hits a target VMAF, capturing quality without overspending bits. VMAF turns the fuzzy question of "does this look good?" into a measurable number that drives ladder design and automated QC.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
VMAF only pays off when the target score is tied to real ladder and bitrate decisions, not reported after the fact. We wrote seven volumes covering the full video stack, and for VMAF start with Volume 1: quality metrics, rate control, per-title encoding and codec trade-offs. Volumes 2 and 3 then follow the encode through packaging, ABR and the CDN bill. Written by Nikolay Sapunov, CEO at Fora Soft.
Picking a VMAF target that cuts bitrate without viewers noticing?
Send us your catalog profile and we will propose VMAF targets and a per-title ladder with the expected bitrate and CDN savings. 250+ video products shipped since 2005.
