Audio-Reactive Visuals — Per-Band, Per-Parameter, Per-Beat.

"Audio-reactive" is a vague label that covers everything from a single brightness pump on the kick to a full modulation graph routing eight frequency bands into dozens of shader parameters. Most VJ apps land near the bottom of that range. A handful land near the top. The gap matters because it's the difference between visuals that move and visuals that perform.

Below is the practical ladder of audio-reactive techniques, ordered roughly from simplest to most expressive. Every rung above the one before it gives a VJ more control over what the visual says about the music — not just that there's music playing, but which part of the music is driving which part of the picture.

The signal flow is straightforward and the architectural choices are deliberate. Audio comes in at 44.1 or 48 kHz from any macOS audio input — system audio, USB audio interface, Aggregate Device, or a DJ controller. It's routed to a dedicated analyzer thread that runs an FFT every audio buffer, not every video frame. That's the first reason RenderWave's audio reactivity is tighter than tools that analyze on the render thread: the analyzer never blocks the GPU and the GPU never blocks the analyzer.

The FFT output is binned into eight or more frequency bands with adjustable crossover points. Each band feeds an envelope follower with independent attack and release times — that's where you tune the response from snappy (techno kick) to glassy (ambient pad). The follower output is the modulation source the shader pipeline actually reads.

On top of the continuous bands, an onset detector watches for percussive transients. The kick detector listens to the sub-bass envelope's derivative and fires on the leading edge of a hit, not on the peak — which gives the visual a head start of a few milliseconds against a beat-locked clip launcher. A separate snare detector watches the 1.5–3 kHz band for the characteristic snare burst. Both detectors emit discrete trigger events alongside the continuous modulation values.

The routing layer is where it stops being a technicality and starts being playable. Drag any band to any shader uniform. Set a depth, a curve (linear, exponential, logarithmic, or a custom S-curve), and a smoothing time. Repeat for every uniform on every shader. The mapping is saved per shader, per scene, per show — and it cross-fades cleanly when you swap between presets, so the audio routing never resets in the middle of a set.

Tempo is handled separately. The BPM estimator runs continuously, and Ableton Link is exposed for any host on the network — Live, a hardware sampler, another VJ rig. Where most VJ tools treat tempo and audio modulation as the same thing, RenderWave keeps them separate because they answer different questions: tempo says when, audio modulation says how loud / how bright / how warped right now.

The shader pipeline reads the band values, onset triggers, and tempo phase as a single uniform buffer at the start of every render pass. There's no CPU round-trip, no audio-to-render thread sync barrier, and no per-shader allocation. That's why the routing layer can stack across eight effects without spending the GPU budget on bookkeeping.

Audio-to-visual latency is the gap between a kick hitting the speaker and the visual responding to it. Below about 30 ms, audiences read the visual as locked to the music. Above 50 ms, it reads as delayed — even if your sync to BPM is perfect, the picture lags the room. Beyond 100 ms, it's actively distracting.

RenderWave targets under 16 ms total audio-to-visual latency on Apple Silicon M2 or newer. That budget breaks down across three stages: audio buffer arrival (driven by the audio interface buffer size, typically 5–10 ms), FFT analysis and envelope follow (sub-millisecond on Apple Silicon's Accelerate / vDSP path), and shader pipeline dispatch to the first visible frame (one to two frame intervals at 60 Hz, so 16–33 ms in the worst case — and well under one frame on ProMotion 120 Hz displays).

Cross-platform VJ tools running over OpenGL or Vulkan-to-Metal translation layers tend to land in the 30–60 ms range because every layer adds a buffer. RenderWave skips all of that by running direct-to-Metal compute with a unified-memory zero-copy texture path.

Ableton Link is a network protocol for sharing tempo and bar phase between music apps on the same LAN. RenderWave participates as a peer — discover Ableton Live, a hardware sampler, another VJ rig, or any Link-capable app on the network and they all share the same beat and the same bar position without timecode cabling, MTC, or MIDI clock drift.

For audio reactivity, Link gives you tempo-locked transitions on top of FFT-driven modulation. The kick still drives the strobe; the bar phase still snaps your shader swap to a clean musical boundary. Tempo and audio are independent inputs into the same modulation graph — change one without breaking the other.

Primary reference: github.com/Ableton/link.