VMAF
An Emmy-winning perceptual video quality assessment algorithm that fuses multiple metrics to predict human visual quality.
What is VMAF?
VMAF is an open-source perceptual video quality assessment algorithm developed by Netflix that predicts human visual quality by fusing multiple elementary metrics. It solves the problem of accurately measuring video quality as perceived by viewers, which is critical for optimizing video encoding, streaming, and quality assurance workflows. The project includes a C library, Python tools, and integrations with FFmpeg and Docker.
Video engineers, researchers, and developers working on video encoding, streaming optimization, quality assessment, and media processing pipelines who need accurate perceptual quality metrics.
Developers choose VMAF because it provides a more human-aligned quality metric than traditional tools like PSNR or SSIM, is backed by Netflix's extensive research, and offers high performance through optimized implementations and flexible integration options.
Overview
Perceptual video quality assessment based on multi-method fusion.
Use Cases
Best For
- Optimizing video encoding settings for streaming services
- Comparing video codec performance in subjective quality
- Integrating perceptual quality metrics into automated video processing pipelines
- Training custom video quality models for specific content types
- Detecting banding artifacts with the included CAMBI metric
- Conducting video quality research and academic studies
Not Ideal For
- Real-time video processing systems needing instant quality feedback
- Projects where traditional metrics like PSNR are sufficient and simplicity is preferred
- Organizations with strict licensing policies that avoid patent grants
Pros & Cons
Pros
Fuses multiple elementary metrics like VIF and DLM to predict subjective video quality, outperforming PSNR and SSIM in correlating with human perception, as highlighted in Netflix's tech blogs.
Implements fixed-point and x86 SIMD (AVX2, AVX-512) for 2x speed improvements over previous versions, making it efficient for batch processing of video encodes.
Available as a C library, Python library, command-line tool, FFmpeg filter, and Docker image, supporting diverse deployment scenarios from research to production pipelines.
Includes Python tools to train and validate custom VMAF models tailored to specific content types, enabling enhanced accuracy for niche use cases like animation or low-light video.
Cons
v3.0.0 removed deprecated APIs, indicating a history of breaking changes that can disrupt existing integrations and require code updates, as noted in the release notes.
Building on Windows requires following specific instructions, and cross-platform deployment with dependencies like Python and FFmpeg adds overhead compared to plug-and-play tools.
Despite optimizations, VMAF can be computationally heavy for high-resolution or real-time video streams, limiting its use in low-resource environments or edge devices.
Frequently Asked Questions
Related Projects
QCToolsQCTools (Quality Control Tools for Video Preservation) is a free and open source software tool that helps users analyze and understand their digitized video files through use of audiovisual analytics and filtering. QCTools is funded by the National Endowment for the Humanities and the Knight Foundation, and is developed by the Bay Area Video Coalition.
PhotonPhoton is a Java implementation of the Interoperable Master Format (IMF) standard. IMF is a SMPTE standard whose core constraints are defined in the specification st2067-2:2013
Pi Audio MonitorAudio Monitoring Software designed to run on a Raspberry pi with touchscreen
BeaqleJS*BeaqleJS* provides a framework to create browser based listening tests and is purely based on open web standards like HTML5 and Javascript.
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.