Multiband Compressor — v1.1 User Guide

What this does

Multiband Compressor v1.1 splits the audio into three frequency bands (low, mid, high) using Hann band‑pass filters, compresses each band independently with its own threshold, ratio, and makeup gain, and then recombines them. The key improvements over v1.0 are:

• Perfect reconstruction – low and high are filtered; mid = original – low – high. This ensures that low + mid + high = original exactly, eliminating the phasy crossover‑boost artefacts of v1.0.
• Envelope‑following compression – gain reduction is computed from the band’s RMS level (measured at 1 kHz), smoothed with attack/release, resampled, and applied as a gain envelope. v1.0 used a sample‑by‑sample Formula that distorted the waveform.
• Attack, release, and makeup gain are now properly implemented and effective.
• Solo mode allows listening to individual bands for fine‑tuning.

Quick start

1. In Praat, select exactly one Sound object.
2. Run script… → Multiband_Compressor.praat.
3. Choose a Preset:
   • Master Bus, Vocal Polish, Bass Control, Brighten & Punch, De‑Harsh, Broadcast Ready, Lo‑Fi Crush
4. For custom mode (preset = Custom), adjust parameters as desired:
   • Crossover frequencies: Low_mid and Mid_high (Hz).
   • Per‑band threshold, ratio, makeup gain.
   • Global attack and release (ms).
5. Optionally choose a Solo band to listen to only low, mid, or high.
6. Click OK. The script splits the sound, compresses each band, recombines, and imports the result as originalname_multiband (or originalname_BAND_solo).

The 7 presets (+ Custom)

Perfect reconstruction filter bank

The Hann band filters are implemented with Praat’s Filter (pass Hann band) using a 100 Hz smoothing width, which gives steep but smooth transitions. The subtraction method ensures perfect reconstruction even with these transitional slopes.

Envelope‑following compression

Each band is compressed using the following steps (no sample‑by‑sample distortion):

1. Level detection at 1 kHz: Convert the band to an Intensity object (time step = 1 ms). This gives a dB‑domain envelope of the band’s RMS level.
2. Gain reduction calculation: For each 1 ms frame, compute required gain reduction in dB:
   if level_dB > threshold_dB: gr_dB = (threshold_dB - level_dB) × (1 – 1/ratio)
   (negative values = attenuation). Convert to linear gain: gain = 10^(gr_dB/20).
3. Attack/release smoothing: Apply exponential smoothing to the gain envelope at 1 kHz using per‑band attack/release coefficients (α = 1 – exp(–4.6 / samples) for a 99 % transition).
4. Resample: Upsample the 1 kHz gain envelope to the audio sample rate using Praat’s sinc resampler (C‑level, fast).
5. Apply: Multiply the band signal by the gain envelope, then by makeup gain (linear).

This approach ensures smooth, click‑free gain changes and preserves the waveform shape – only the amplitude envelope is modified.

Parameters & defaults

Crossover frequencies

Per‑band controls

Global & output

Visualization (Praat picture)

When Visualize = 1, the script draws:

• Input waveform (grey).
• LOW band (red), MID band (green), HIGH band (blue) after compression.
• Output waveform (purple).
• Band info panel – colour‑coded rectangles showing threshold and ratio for each band.

FAQ / troubleshooting