Enable Variance-Based AQ in SVT-AV1

This article provides a straightforward guide on how to enable variance-based adaptive quantization (AQ) when encoding video using the SVT-AV1 encoder. It covers the specific command-line parameters required for both the native SVT-AV1 standalone encoder (SvtAv1EncApp) and the FFmpeg integration, allowing you to improve visual quality by distributing bits more efficiently across different frame textures.

Understanding the Parameter

Adaptive Quantization (AQ) adjusts the quantization parameter (QP) locally within a frame based on the complexity of the image. In SVT-AV1, variance-based AQ analyzes the spatial variance of a block to determine whether it contains flat, low-contrast textures or highly detailed, high-contrast textures. This helps prevent blockiness and banding in smooth areas like skies or flat walls.

To enable variance-based adaptive quantization, you must set the AQ mode parameter to 1.

Enabling AQ in the Native SVT-AV1 Encoder

When using the standalone SvtAv1EncApp executable, you enable variance-based AQ by passing the --aq-mode flag followed by the value 1.

Below is an example of the command-line usage:

SvtAv1EncApp -i input.yuv -w 1920 -h 1080 --preset 4 --crf 26 --aq-mode 1 -b output.ivf

Enabling AQ in FFmpeg

If you are invoking libsvtav1 through FFmpeg, you must pass the parameter using the -svtav1-params option. Multiple parameters inside -svtav1-params are separated by colons.

Below is an example of the FFmpeg command-line usage:

ffmpeg -i input.mp4 -c:v libsvtav1 -crf 26 -preset 4 -svtav1-params aq-mode=1 output.mkv

Verifying the Settings

• aq-mode 0: Disables adaptive quantization entirely.
• aq-mode 1: Enables variance-based adaptive quantization (recommended for most standard content to improve subjective visual quality).
• aq-mode 2: Enables delta-q adaptive quantization (an alternative method that utilizes delta-QP to optimize rate-distortion performance).