← sources

[FOSS] I cut my jellyfin library ~50% with a self-hosted, ultra-optimzied AV1 converter (Web UI, Docker, PGO, LTO)

source: reddit (reddit:selfhosted)published 2026-10-30 13:52ingested 2026-10-31 02:37

retained text

[Screenshot](https://preview.redd.it/3hugxjjpzlyh1.png?width=2880&format=png&auto=webp&s=1c177ae23498a7bccf1462047dbb3375f9245ecb)

I'm the author - free + MIT.

My library was mostly MKVs and MP4s, which eat storage like crazy, depending on the source. So I created archive-video-av1. It watches a folder and converts files to AV1 (SVT-AV1) in place, with a queue you can follow in the browser and clustering, if needed.

On speed: in my test with exactly the same encode settings, my source build was about 30% faster than the already quite optimized ffmpeg + SVT-AV1 from CachyOS on a Zen5 CPU. The prebuilt Docker image is a generic portable build, so it is slower. For maximum efficiency, make sure you compile yourself, with your own (30s-1min) PGO training source material and the easy to use "make" commands.

What it does:

* Videos -> AV1 in MKV (`_conv.mkv` next to source), Opus audio with two-pass loudnorm
* Presets for Default, Animated, Grainy / Very Grainy, plus custom presets with import/export
* Batch queue with history, filters, retry, live progress via WebSocket
* Skips files already on AV1, never overwrites, keeps queue/history across restarts
* Optional LAN-distributed queue if you have several machines

Honestly, just run it with the built-in defaults. I hope you like it and feel free to leave me some feedback. Source + docs: [https://github.com/fabianwimberger/archive-video-av1](https://github.com/fabianwimberger/archive-video-av1)

AI Involvement: Made with AI assistance

Update: fixed typos. If you already use Tdarr, stick with it.

document fetched 2026-10-31 00:29 · canonical https://reddit.com/r/selfhosted/comments/1xjfypz/foss_i_cut_my_jellyfin_library_50_with_a