Hylas — AI music visualiser system

MILKWAVE SPECTRAFORGE Real-time music visualisation, driven by AI

MILKWAVE is a music visualiser that listens to whatever is playing and renders a live, beat-reactive light show — pulling from a library of 148,000 hand-crafted visual presets and using AI to pick the right one for the moment.

What is it How it works Technical ›

148k

Visual presets

Real-time

Beat response

Preset selection

Any

Audio source

01 What is it

The visualiser

MILKWAVE

MILKWAVE is a real-time music visualiser for Windows. Point it at any audio source — a playlist, a live DJ mix, system audio — and it renders a full-screen animated light show that reacts to every beat, bass hit, and frequency shift.

It uses the MilkDrop visualisation engine, a legendary piece of software originally released in 2001 that turns mathematical GLSL shader code into flowing, organic visuals. MILKWAVE brings that engine into the modern era with an AI layer that makes preset selection intelligent rather than random.

The preset engine

SPECTRAFORGE

SPECTRAFORGE is the component that manages MILKWAVE's library of 148,000 visual presets — each one a unique, hand-crafted shader written by the MilkDrop community over the past two decades.

Its job is to classify, tag, and serve those presets intelligently. Rather than cycling through them randomly, SPECTRAFORGE understands what each preset looks like — whether it's geometric, fluid, strobing, or calm — so the AI layer can match the visual mood to the music.

02 How it works

FROM SOUND TO SCREEN

Audio comes in, visuals come out — but there's quite a bit happening in between.

LISTEN

Captures whatever audio is playing on your system — music, streams, anything. No need to plug in cables or change your setup.

ANALYSE

The audio is broken into frequency bands — bass, mids, treble — and analysed for rhythm, tempo, and transient hits many times per second.

DECIDE

An AI orchestrator reads the audio analysis and picks a preset from the library that matches the current musical mood and energy level.

RENDER

The chosen preset runs as a live GLSL shader. Every bass hit, beat, and melody is fed directly into the visual — warping, pulsing, and flowing in real time.

TRANSITION

As the music evolves, the AI switches presets at natural breaks — blending between visuals smoothly rather than cutting abruptly.

03 Spectraforge

THE PRESET LIBRARY

148,000 presets is an extraordinary resource — but only if you can find the right one. SPECTRAFORGE makes the library usable.

Classify

TAGGING

Every preset is analysed and tagged by visual character — geometric shapes, fluid motion, strobing patterns, dark atmospheres, bright energy. The AI learns what each one looks like so it can be retrieved by mood.

Index

FAST SEARCH

The full 148k library is indexed in under a second. When the AI needs a "deep bass, dark, geometric" preset mid-track, the answer comes back instantly — no lag, no interruption.

Match

MOOD MAPPING

Audio energy maps to visual character. Aggressive, high-BPM music surfaces sharp, strobing presets. Ambient, slow passages pull from fluid, atmospheric ones. The match is automatic and continuous.

History

COMMUNITY LIBRARY

The preset library is sourced from the MilkDrop community — two decades of creative work by musicians, visualists, and demoscene artists. SPECTRAFORGE curates and navigates this archive.