Audio Filter

Audio Filter sits anywhere in an audio chain: source → filters → Audio Output. Connecting the Audio input splices the node into the shared AVAudioEngine graph — the cables are the engine wiring, everything renders on one clock, and the node can be inserted or removed mid-playback without a click.

Internally the node is a single-band AVAudioUnitEQ. The Type port selects the band's filter shape (resonant low-pass, resonant high-pass, band-pass, notch, low shelf, or high shelf). Freq positions the filter in the spectrum. Resonance sets the bandwidth in octaves — narrower bandwidth means a sharper, more resonant response around the cutoff, the sound of an emphasized filter sweep. Gain applies only to the two shelf types, boosting or cutting everything below (Low Shelf) or above (High Shelf) the corner frequency.

The Audio output re-emits the filtered signal as a new engine-backed snapshot, so filters chain indefinitely and an Audio FFT can watch the signal anywhere in the chain — tap after a low-pass and the FFT sees only the bass. The Level output is a peak-decay RMS of the filtered signal (instant rise, musical fall). Enable = false engages the audio unit's clean bypass: audio still passes, unprocessed.

Filter types (Type index):

Outputs: Audio (the filtered signal, chainable) and Level (peak-decay RMS of the filtered signal).

Musical guidance: a high-pass around 100–200 Hz cleans mud from a mix; a low-pass swept from 20000 down to a few hundred Hz is the classic breakdown filter; a band-pass at 500–2000 Hz with tight resonance (0.1–0.3 octaves) gives a telephone/radio voice; a notch is surgical removal of a single offending frequency.

• Bass-only visuals: Audio Input → Audio Filter (Low Pass, Freq ≈ 150) → Audio Output, with an Audio FFT tapped on the filter's Audio output — the spectrum and beat detection now see only the low end.
• LFO filter sweep: Audio Player → Audio Filter (Low Pass) → Audio Output, with Freq cabled to an LFO scaled to roughly 200…8000 Hz — an automatic, tempo-free wah sweep.
• DJ filter kill on a fader: Audio Player → Audio Filter → Audio Output, Freq wired to a UI slider; sweep down for a low-pass kill (Type 0) or up for a high-pass kill (Type 1) during transitions.
• Telephone voice: Audio Player (voice) → Audio Filter (Band Pass, Freq 1200, Resonance 0.3) → Audio Distortion (Cellphone preset) → Audio Output.
• Parallel band split: fan one Audio Player into two chains — Audio Filter (Low Pass, 200 Hz) and Audio Filter (High Pass, 2000 Hz) — process each branch differently (e.g. reverb only on the highs) and route both to the Audio Output; the engine supports parallel branches natively.
• Hum removal: Audio Input → Audio Filter (Notch, Freq 50 or 60, Resonance 0.05) → Audio Output — a tight notch on mains hum.

• Resonance is bandwidth in octaves, not a Q knob: lower values mean sharper, more resonant filtering. For screaming sweep emphasis go toward 0.05–0.2; for gentle tone shaping stay near 1–2.
• Gain does nothing on Low Pass, High Pass, Band Pass, or Notch — it is only honored by the two shelf types. For overall level, chain an Audio Gain node instead.
• Sweeping Freq with a logarithmic-feeling control (or an exponential mapping via Math nodes) sounds far more natural than a linear sweep, because pitch perception is logarithmic.
• Place the filter before a delay or reverb to color the whole effect, or after to filter the wet tail too — order changes the sound.
• Enable = false is a true clean bypass, ideal for A/B-ing the filter on a toggle; the chain and its analysis taps stay live either way.
• Live microphone or loopback input runs through the same engine via the realtime source bridge, so all six filter types work on live audio identically.