How SVT-AV1 2.0 Changes Default Preset Speeds

The release of SVT-AV1 version 2.0 introduced significant structural changes to the encoder’s preset system, altering how speed and quality trade-offs are managed. This article explains how the transition to version 2.0 reorganized the preset scale, shifted the default encoding behaviors, and impacted performance benchmarks for users upgrading from older versions.

The Compression of the Preset Scale

Prior to version 2.0, SVT-AV1 utilized a preset scale ranging from 0 to 13, where lower numbers represented slower, higher-quality encodes, and higher numbers offered faster processing at the expense of compression efficiency.

In SVT-AV1 2.0, the development team overhauled this system to compress the scale and make room for future optimizations. The scale was reduced to 0 through 12, completely removing Preset 13. Because of this consolidation, the entire range of presets was remapped. Users transitioning to version 2.0 cannot expect a preset number (such as Preset 8) to behave exactly the same as it did in version 1.x.

Shifted Quality and Speed Dynamics

The remapping of the presets changed the actual encoding speed and quality characteristics of almost every level:

• Lower Presets (0 to 6): These presets received substantial optimizations to improve their quality-to-performance ratio. While some of these slower presets may take slightly longer to render compared to their v1.x counterparts, they deliver significantly better compression density.
• Higher Presets (7 to 12): The faster presets were retuned to better handle real-time encoding and high-throughput workloads, absorbing the performance characteristics of the retired Preset 13.

The New Default Preset Behavior

In many integration tools and libraries (such as FFmpeg), the default preset behavior has shifted to align with the new v2.0 standards.

The library now establishes Preset 6 as the standardized default for general encoding. Preset 6 represents the “sweet spot” for offline encoding, offering an optimal balance between CPU utilization and high-fidelity video compression. For users who previously relied on default settings in older versions of SVT-AV1 (which often leaned toward faster, lower-quality defaults like Preset 8 or 10 in certain frameworks), the transition to version 2.0’s defaults may result in longer encoding times but will yield vastly superior file sizes and visual quality.

What This Means for Users

If you are upgrading an existing video processing pipeline to SVT-AV1 2.0, you should not copy over your old configuration parameters directly. Instead, you should re-evaluate your command-line arguments:

1. Re-test your presets: Run test encodes to find the new numbering equivalent for your speed-to-quality threshold.
2. Adjust for the default shift: If you do not specify a preset in your command line, expect the encoder to run slower than before, as it now defaults to the highly-optimized Preset 6.
3. Use Preset 8 or higher for speed: If encoding speed is your primary constraint, start your testing at Preset 8 and move upward to find the correct balance for your hardware.