mcpbeat

2000s Visualization Expert

curiositech/2000s-visualization-expert

Expert in 2000s-era music visualization (Milkdrop, AVS, Geiss) and modern WebGL implementations. Specializes in Butterchurn integration, Web Audio API AnalyserNode FFT data, GLSL shaders for audio-reactive visuals, and psychedelic generative art. Activate on "Milkdrop", "music visualization", "WebGL visualizer", "Butterchurn", "audio reactive", "FFT visualization", "spectrum analyzer". NOT for simple bar charts/waveforms (use basic canvas), video editing, or non-audio visuals.

7k tokens
context cost
the whole folder, loaded on every use
5
files
instructions only
0
copies elsewhere
how many repositories repackaged it
177
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/curiositech/some_claude_skills --skill 2000s-visualization-expert

The instruction itself

18 sections, as written by the author

2000s Music Visualization Expert

Expert in recreating the legendary 2000s music visualization era - Milkdrop, AVS, Geiss - using modern WebGL and Web Audio APIs.

When to Use

Use for:

  • Implementing Milkdrop-style psychedelic visualizations
  • Butterchurn library integration (WebGL Milkdrop)
  • Web Audio API AnalyserNode FFT/waveform extraction
  • GLSL fragment shaders for audio-reactive effects
  • Full-screen immersive music experiences
  • Real-time beat detection and audio analysis
  • Preset systems and visualization transitions

NOT for:

  • Simple spectrum bar charts (use Canvas 2D)
  • Static audio waveform displays
  • Video editing or processing
  • Non-audio generative art
  • Audio playback/streaming issues (use audio-engineer skills)

The Golden Era (Summary)

| Era | Key Software | Innovation |

|-----|--------------|------------|

| 1998-2000 | Geiss | Simple plasma effects, DirectX |

| 2001-2007 | Milkdrop 1 & 2 | Per-pixel equations, preset system |

| 2007-2015 | Decline | Streaming services rise |

| 2018-Present | Butterchurn | WebGL renaissance |

Milkdrop's magic: Layering simple effects - blur, zoom, rotation, color shift - with audio-reactive parameters.

→ See references/butterchurn-guide.md for full history and integration.

Core Technologies

Butterchurn (WebGL Milkdrop)

  • 1.7k GitHub stars, MIT licensed
  • Full preset compatibility with original Milkdrop
  • npm: butterchurn, butterchurn-presets
import butterchurn from 'butterchurn';
const visualizer = butterchurn.createVisualizer(audioContext, canvas, {
  width: window.innerWidth,
  height: window.innerHeight,
  pixelRatio: window.devicePixelRatio || 1,
});
visualizer.connectAudio(audioNode);
visualizer.loadPreset(preset, 2.0);  // 2s blend

Web Audio API FFT

const analyser = audioContext.createAnalyser();
analyser.fftSize = 2048;
analyser.smoothingTimeConstant = 0.8;
const frequencyData = new Uint8Array(analyser.frequencyBinCount);
analyser.getByteFrequencyData(frequencyData);

Critical: FFT bins are linear but hearing is logarithmic! Use logarithmic mapping.

→ See references/web-audio-fft.md for frequency band extraction.

GLSL Shaders

Pass audio data as 1D texture, use uniforms for bass/mid/treble:

uniform float u_bass;
float glow = smoothstep(0.5 - u_bass * 0.3, 0.0, dist);

→ See references/glsl-shaders.md for complete patterns.

Anti-Patterns to Avoid

1. Ignoring AudioContext State

What it looks like: Visualization silently fails

Why it's wrong: AudioContext starts suspended, needs user interaction

Fix: Resume on click: await audioContext.resume()

2. Linear Frequency Display

What it looks like: Bass dominates, treble invisible

Why it's wrong: FFT bins are linear; first 100 bins might be 0-2kHz

Fix: Use logarithmic bin mapping (code in references)

3. No Smoothing

What it looks like: Jittery, seizure-inducing visuals

Why it's wrong: Raw FFT data is noisy frame-to-frame

Fix: analyserNode.smoothingTimeConstant = 0.7

4. requestAnimationFrame Without Cleanup

What it looks like: Memory leaks, multiple render loops

Fix: Store animation ID, call cancelAnimationFrame on unmount

5. Hardcoded Canvas Size

What it looks like: Blurry on retina, wrong aspect ratio

Fix: Multiply by devicePixelRatio, handle resize events

6. Blocking Main Thread

What it looks like: Choppy audio, dropped frames

Why it's wrong: Heavy shader compilation on UI thread

Fix: Compile shaders during loading, not during playback

Preset Recommendations

Psychedelic/Trippy:

  • Flexi, martin + geiss - dedicated to the sherwin maxawow
  • Rovastar - Fractopia

Smooth/Chill:

  • Flexi - predator-prey-spirals
  • Geiss - Cosmic Strings 2

High Energy:

  • Flexi + Martin - disconnected
  • shifter - tumbling cubes

Integration Checklist

  • [ ] AudioContext created and resumed on user interaction
  • [ ] AnalyserNode connected to audio source
  • [ ] Canvas sized correctly (account for devicePixelRatio)
  • [ ] Render loop with requestAnimationFrame
  • [ ] Cleanup on unmount (cancelAnimationFrame)
  • [ ] Preset loading with blend time
  • [ ] Resize handling
  • [ ] Full-screen support with ESC to exit
  • [ ] Track info overlay (z-index above canvas)
  • [ ] Cursor hiding after inactivity

Performance Tips

  • Lower texture ratio for older GPUs: textureRatio: 0.5
  • Reduce fftSize if not needed: 512 or 1024 vs 2048
  • Use will-change: transform on canvas
  • Avoid DOM updates during render loop
  • Profile with Chrome DevTools GPU timeline

References

references/butterchurn-guide.md - Complete Butterchurn integration

references/web-audio-fft.md - FFT extraction and frequency analysis

references/glsl-shaders.md - Audio-reactive shader patterns


This skill covers: Butterchurn/Milkdrop | Web Audio FFT | GLSL shaders | Full-screen visualization | Audio-reactive art

How to use it

Copy the folder

Take curiositech/2000s-visualization-expert from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.