# Auto-Boost Av1an Portable

Canonical URL: https://encode.wiki/tools/utilities/auto-boost-av1an/
Content version: ca42e61ede0a24f789d7287a0001244ad22f4abf
Status: current
License: CC BY-SA 4.0
Checksum (SHA-256): 2ee2f753e457b1b854a31694d37f98917bcf6dc162e4c7f1b7750fadfdc67f16

> **Repository:** [LastBreeze/Auto-Boost-Av1an-portable](https://github.com/LastBreeze/Auto-Boost-Av1an-portable). Fully automated AV1 encoding workflow for Windows. Drop .mkv files, double-click a batch .bat, done. Uses av1an with Auto-Boost for consistent visual-metric quality. Supports any quarter-step CRF SVT-AV1 fork, zones.txt, and multiple worke…. Latest release: [3.4.7](https://github.com/LastBreeze/Auto-Boost-Av1an-portable/releases/tag/3.4.7), Sep 29, 2026. Last commit: Aug 22, 2026. License: GPL-3.0. Stars: 28.

Auto-Boost is a technique for more consistent quality from [Av1an](/tools/utilities/av1an/) encodes. A fast first encode is made at the base CRF, every scene is scored with [SSIMULACRA2](/reference/metrics/ssimulacra2/) (and in later versions XPSNR), and the CRF of each scene is then raised or lowered so that its score moves toward the average. The adjusted values are written as an Av1an `zones.txt` file and used for the final encode. Hard scenes get more bits; easy scenes give some back.

The scripts were written by nekotrix and others in the [auto-boost-algorithm](https://github.com/nekotrix/auto-boost-algorithm) repository; the current script there, Auto-Boost-Essential, pairs with the [SVT-AV1-Essential](/tools/software-encoders/svt-av1-essential/) fork. Like Av1an's own target-quality mode, it is slower than a single encode and spends bits by metric rather than by eye.

## Auto-Boost Av1an Portable

[Auto-Boost-Av1an-portable](https://github.com/LastBreeze/Auto-Boost-Av1an-portable) packages this workflow for Windows users who do not want to set up Av1an, VapourSynth and Python themselves. It is maintained (3.4.7, September 2026) and GPL-3.0 licensed. A [Linux fork](https://github.com/abdalrahmanx9/Auto-Boost-Av1an-Linux) exists separately.

1. Download and extract the release archive.
2. Put your `.mkv`, `.mp4` or `.m2ts` files in `video-input`.
3. Double-click one of the included batch files, such as `autoboost-anime-crf30.bat` or `autoboost-liveaction-crf30.bat`, or run `bat-builder.bat` to generate one for your content and hardware.
4. Collect the results from `video-output`. The original audio and subtitle tracks are remuxed back in.

The `autoboost-*` scripts run the two-pass Auto-Boost process; the `av1an-batch-*` scripts do a single Av1an encode. You choose among bundled SVT-AV1 forks (`5fish`, `essential`, `hdr`, or your own build in `custom`). The first run benchmarks your machine to pick a worker count and whether SSIMULACRA2 can run on the GPU through Vship; delete `tools\workercount-config.txt` to repeat it. Interrupted batches resume on the next run, except that a file's final pass restarts from the beginning.