Skip to main content

VMAF (Video Multimethod Assessment Fusion)

VMAF (Video Multimethod Assessment Fusion) is AI-powered quality assessment algorithm that combines multiple video analysis methods to evaluate video quality across diverse content types, providing scores that match human perception by comparing processed videos against their original versions.

Example: A video streaming service wants to ensure that the animated background on its login page looks consistently high-quality for all users. The quality assurance team would use VMAF to test the video. They would:

  1. Take a pristine, uncompressed version of the login animation as the reference video.
  2. Run tests by compressing the video to different bitrates and resolutions.
  3. Use the VMAF tool to compare each compressed version against the original reference video.

The VMAF score would provide an objective measure of how much visual quality was lost in each compressed version. This allows the team to find the optimal balance between a low file size (to save bandwidth and ensure fast loading) and high visual quality (to provide an excellent user experience).