Skip to main content
ffmpeg-quality-metrics calculates video quality metrics using FFmpeg. It supports SSIM, PSNR, VMAF, and VIF metrics for comparing video quality.

​
Installation

​
Run directly

nix run github:spotdemo4/nur#ffmpeg-quality-metrics

​
Add to flake

devShells.default = pkgs.mkShell {
  packages = with pkgs.trev; [
    ffmpeg-quality-metrics
  ];
};

​
Usage

​
Compare two videos

Compare a distorted video against a reference: 
ffmpeg-quality-metrics reference.mp4 distorted.mp4

​
Calculate VMAF score

VMAF (Video Multimethod Assessment Fusion) is a perceptual video quality metric: 
ffmpeg-quality-metrics reference.mp4 distorted.mp4 -m vmaf

​
Calculate multiple metrics

Calculate PSNR, SSIM, and VMAF: 
ffmpeg-quality-metrics reference.mp4 distorted.mp4 -m psnr ssim vmaf

​
Export results to JSON

ffmpeg-quality-metrics reference.mp4 distorted.mp4 -of json > metrics.json

​
Supported metrics

Measures the ratio between the maximum possible signal power and the power of corrupting noise. Higher values indicate better quality. Typical values range from 30-50 dB.
Measures the perceived quality based on structural information. Values range from 0 to 1, where 1 indicates perfect similarity.
Netflix’s perceptual video quality metric. Scores range from 0-100, with higher values indicating better quality. Scores above 95 are considered excellent.
Measures the amount of information that could theoretically be extracted from the reference image. Values range from 0 to 1.

​
Common use cases

​
Evaluate encoding quality

Compare original video with encoded version: 
ffmpeg-quality-metrics original.mp4 encoded.mp4 -m vmaf psnr

​
Test different encoding settings

Batch compare multiple encoded files: 
for file in encoded_*.mp4; do
  echo "Testing $file"
  ffmpeg-quality-metrics original.mp4 "$file" -m vmaf
done

​
Continuous quality monitoring

Integrate into CI/CD pipelines: 
ffmpeg-quality-metrics reference.mp4 output.mp4 -m vmaf -of json | \
  jq '.vmaf.mean' | \
  awk '{if ($1 < 80) exit 1}'
VMAF scores correlate well with human perception. Target VMAF scores above 80 for good quality, and above 95 for excellent quality.

Build docs developers (and LLMs) love

Get started for freeTalk to us