One person, one engine, one laptop.

Sainted Word Records is a single-person audio-reactive video studio. I built the engine because the existing tools split into three bad buckets: template mush, AI mush, and pro VJ rigs that cost more than my first car.

Why a fourth category

The honest answer: I wanted to make music videos for my own songs without paying €600/year for Resolume or learning TouchDesigner for three months. And I wanted the engine to learn from what I actually use — not the other way around.

SWR is a browser-native, library-driven, audio-reactive video engine. Drop a song, drop a library of clips, get a layered composition that breathes with the music. The engine analyses the audio (BPM, key, beats, chromagram) and renders a real-time reactive composition. Output as WebM at 1080p.

What "self-evolving" means here

The engine ships with 23 personas (10 original + 5 generative + 5 MORPHA + 3 train-stages), 14 WebGL FX, and 20 stackable CSS filters. That's the seed corpus. On top of that, a daily pipeline generates 2 new preset specs — one per day, committed to the repo and shipped to the live engine. After 3 uses of a single persona, the engine itself generates a personalized variant tuned to your song's actual BPM and key.

The idea isn't novelty for its own sake — it's that the engine should converge toward what each user actually wants from it, without uploading usage data to anyone.

Principles

How to support

Three ways, none required:

  1. Use it. Open the engine, make something, post it.
  2. Tell someone. One sentence is fine.
  3. Order a render or buy a Personal Tier license if you want to support the project financially.

Kai Djuric · Belgrade, Serbia · kai [at] saintedwordrecords [dot] com
GitHub · Source · Changelog Intro About · Press kit · The Atlas