How to Integrate librav1e into FFmpeg

This guide provides a comprehensive overview of integrating and using the rav1e AV1 encoder (via the librav1e library) within FFmpeg for efficient video processing. You will learn how to obtain or compile an FFmpeg binary with librav1e support, write the correct command-line syntax for encoding, and adjust key parameters such as speed presets and constant quality to optimize your AV1 video compression workflow.

Step 1: Enable librav1e in FFmpeg

To process videos using librav1e, your FFmpeg installation must be compiled with support for it. Because librav1e is licensed under the BSD 2-Clause license, it can be built into both GPL and non-GPL builds of FFmpeg.

If you are compiling FFmpeg from source, you must first install Rust, cargo, and the rav1e library on your system. Once those dependencies are met, configure your FFmpeg build by appending the --enable-librav1e flag:

./configure --enable-gpl --enable-librav1e
make -j$(nproc)
sudo make install

For users who prefer not to compile from source, pre-built static binaries containing librav1e are available through third-party package managers like Homebrew (macOS) or specialized community builds for Windows and Linux.

Step 2: Basic Encoding Syntax

Once you have an FFmpeg build with librav1e enabled, you can specify it as your video encoder using the -c:v librav1e parameter.

The most basic command to transcode a video to AV1 using librav1e is:

ffmpeg -i input.mp4 -c:v librav1e -c:a copy output.mkv

This command takes an input video, encodes the video stream using librav1e, copies the audio stream without re-encoding, and packages the result into a Matroska (.mkv) container.

Step 3: Configuring Quality and Speed

To achieve the best balance between file size, visual quality, and encoding time, you should configure the encoder’s speed presets and quality parameters.

Speed Presets (-preset): rav1e offers speed levels ranging from 0 (slowest, highest compression efficiency) to 10 (fastest, lowest compression efficiency). A preset of 5 or 6 is generally recommended for a good balance of speed and quality.
Quantizer / Quality (-qp): Instead of standard CRF, librav1e uses Quantizer Parameter (QP) values to control quality. The range is 0 to 255. Lower values yield higher quality and larger file sizes. A QP value between 50 and 100 is typically ideal for standard-definition and high-definition video.

An optimized command utilizing these settings looks like this:

ffmpeg -i input.mp4 -c:v librav1e -qp 80 -preset 5 output.mp4

Step 4: Advanced Fine-Tuning

For advanced control, you can pass custom parameters directly to the rav1e library using the -rav1e-params flag. Multiple parameters can be separated by colons.

For example, to set a keyframe interval (gop size) and force low-latency tuning, use the following command:

ffmpeg -i input.mp4 -c:v librav1e -qp 80 -rav1e-params speed=5:low_latency=true -g 240 output.mp4

By leveraging these configuration options, you can successfully integrate librav1e into your FFmpeg pipeline to generate highly optimized AV1 video files.