VMAF is an Emmy-winning perceptual video quality assessment algorithm developed by Netflix. This software package includes a stand-alone C library, libvmaf, and its wrapping Python library. The Python library also provides tools to train and test a custom VMAF model. Also included in libvmaf are implementations of several other metrics: PSNR, PSNR-HVS, SSIM, MS-SSIM and CIEDE2000.
In June 2026, Netflix released a new set of VMAF models (v1), while the previous generation (v0) remains documented. On December 7, 2023, libvmaf v3.0.0 was released, containing optimizations and bug fixes and a full removal of APIs deprecated in v2.0.0. The v2.0.0 release achieved a 2x speed up through fixed-point and x86 SIMD optimization.
Netflix open-sourced CAMBI (Contrast Aware Multiscale Banding Index), its detector for banding artifacts, and changed VMAF's license from Apache 2.0 to BSD+Patent. VMAF is now included as a filter in FFmpeg and can be configured using ./configure --enable-libvmaf. A VMAF Dockerfile generates a docker image from the Python library, and Windows build instructions are also available.
Source: Hacker News · Summarized by HeadlinesBriefing