# Netflix/vmaf

Perceptual video quality assessment based on multi-method fusion.

Repository: https://github.com/Netflix/vmaf
Canonical: https://ross.abutalabs.com/products/vmaf
Language: C
License: NOASSERTION
License Family: other
Last push: 2026-08-14T17:18:22+00:00

## Health v2 (maintenance only)
Score: 91/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 77, longevity 100
- inputs: {"age_days": 3859, "days_push": 19, "days_rel": 75, "gap_med": 78, "n_releases_24m": 2}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5460, forks 830 (observed 2026-08-28T04:09:19.485032+00:00)

## What it is
VMAF is Netflix's Emmy-winning perceptual video quality assessment algorithm, provided as a standalone C library (libvmaf) with a wrapping Python library. It fuses multiple quality metrics into a single score and also includes implementations of PSNR, PSNR-HVS, SSIM, MS-SSIM, CIEDE2000, and the CAMBI banding detector, plus tools for training custom models.

## Use cases
- measure perceptual quality of encoded video against a reference
- compare video codecs with a quality metric
- train a custom video quality model for my content
- detect banding artifacts in video with CAMBI
- compute PSNR, SSIM, or MS-SSIM between two videos
- integrate video quality scoring into an encoding pipeline via ffmpeg

## When to choose
- you need an industry-standard full-reference video quality metric
- you're evaluating or tuning video encoders and codec settings
- you want to train a quality model tailored to your content distribution

## When to avoid
- you need no-reference quality estimation without a source video
- you need audio quality assessment
- you want a simple GUI tool rather than a library/CLI

## Facets
- artifact type: library
- maturity: stable
- function: video-processing, machine-learning, benchmarking
- domain: media, machine-learning
- platform: windows, python, cpp
- tags: video-quality, perceptual-metrics, vmaf, psnr, ssim, ciede2000, cambi, ffmpeg, libvmaf, video, linux, macos, docker

## Member repositories
- Netflix/vmaf (main) score 91

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:19.485032+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:56:46.835090+00:00, confidence not recorded.
  - readme: https://github.com/Netflix/vmaf (fetched 2026-08-28T04:09:19.485032+00:00, sha d1def56f1cb9)
- Data as of 2026-08-30T08:39:29.467469+00:00.