Skip to main content

Compute VMAF

The Compute VMAF task calculates a Video Multi-Method Assessment Fusion (VMAF) score by comparing an encoded video segment against its original reference file to precisely measure perceptual quality.

This task is specifically built to support the FFmpeg Builder: Video Encode Optimized flow element, which relies on VMAF scores to dynamically optimize encoding parameters.

Configuration Settings​

Concurrent Tasks​

  • Description: The maximum number of VMAF calculation tasks that can run simultaneously.
  • Usage: Because metric calculation is exceptionally demanding on system hardware, capping this at 1 or 2 keeps your processing environment stable.

Threads​

  • Description: The number of CPU threads allocated to each individual VMAF calculation task.
  • Usage: Setting a higher thread count speeds up the quality check significantly. Leave it at 0 to automatically inherit your system's global default VMAF thread limits.

Why Use Compute VMAF?​

  • Offload Heavy Calculations: Computing VMAF scores is incredibly resource-heavy and will easily peg your CPU at 100%. By offloading this work to the Forge, you free up your standard FileFlows processing nodes to focus entirely on the actual video encoding without getting bogged down by quality assessment tasks.
  • Scale with Better Hardware: This structure allows you to distribute lightweight video encodes across multiple smaller processing nodes, while routing the compute-heavy VMAF analysis back to a central, high-powered Forge machine capable of chewing through the calculations much faster.