Audio node
Audio FFT
GPU-accelerated FFT of the audio ring buffer via Metal compute shaders (radix-2), with selectable size, window function and smoothing. Produces magnitude data for the Spectrum and Analysis nodes.
Inputs
| Port | Type | Default |
|---|---|---|
| Audio DataThe raw audio to transform. Cable it from Audio Input, Audio Player or the output of an audio effect. | structure | — |
| FFT SizeThe analysis window, from 16 samples up to 64K. Small windows react fast but read coarse in the bass, large ones resolve pitch but smear transients. Out-of-range picks snap to the nearest size. | int | 3 |
| WindowThe window shape applied before the transform: hann, hamming, blackman or flat-top. Anything else falls back to hann. | string | hann |
| ChannelWhich input channel is analysed, counted from 0. Numbers beyond what the source carries fall back to its last channel. | int | 0 |
| FocusConcentrates all bins into a narrower range for far finer bass detail: full range, low, bass or sub focus. Deeper focus costs a longer window, so transients smear; unlisted values fall back to full range. | int | 3 |
| SmoothingHow much of the previous frame is carried into the spectrum. Not clamped; useful roughly 0…0.9, and at 1 the display freezes on old data. | float | 0.5 |
Outputs
| Port | Type |
|---|---|
| FFT Data | structure |
| Bin Count | int |
| Freq Resolution | float |
| Nyquist | float |
What Is an FFT?
Sound is a single wiggling value over time. The FFT (Fast Fourier Transform) answers a different question: "which musical pitches make up this sound, and how loud is each one?" It splits the signal into frequency bins — bass on the left, treble on the right — and gives the strength of each. That's what every spectrum analyzer and "dancing bars" visual is built on.
How It Works
Audio FFT takes the live audio buffer and runs the transform on the GPU (Metal compute shaders), so even large sizes stay real-time. FFT Size trades detail for speed: bigger = finer frequency resolution but a touch more latency. A Window function cleans up the edges of each chunk (reduces artifacts), and Smoothing calms the bars frame-to-frame.
Parameters
| Parameter | What it does |
|---|---|
| FFT Size | 256–4096. Larger = finer pitch detail, slightly more lag. 1024 is a good default. |
| Window | Edge-smoothing function (Hann, etc.) — reduces spectral smearing. |
| Smoothing | How much to ease the spectrum between frames. |
| Channel | Which input channel to analyze. |
Outputs FFT Data (for Spectrum/Analysis) plus Bin Count, Freq Resolution and Nyquist (the highest detectable frequency).
Examples
- Spectrum bars — Audio Input → Audio FFT → Audio Spectrum for a live analyzer.
- Beats & bands — Audio FFT → Audio Analysis for bass/mid/treble and beat detection.
Tips
- Use 512 for snappy, fast-reacting visuals; 2048+ for detailed spectral displays.
- FFT Size should be a power of two (256, 512, 1024…).
Open Audio FFT in the interactive catalog →
Audio FFT is one of 381 nodes in Grafit Pro.
Free download. Every feature for 7 days, no card. Then €99.99 once — or keep playing your projects as a Player.