How to Use FFmpeg firequalizer for Custom EQ Curves

This guide demonstrates how to apply arbitrary frequency response curves to audio streams using FFmpeg's built-in firequalizer (Finite Impulse Response Equalizer) filter. You will learn the two primary methods for shaping the frequency spectrum—using discrete coordinate points with linear interpolation or using continuous mathematical expressions—along with the critical parameters needed to control curve accuracy and sound quality.

Core Concepts of firequalizer

The firequalizer filter shapes audio in the frequency domain using Fast Fourier Transforms (FFT). Unlike traditional parametric equalizers that rely on cascaded IIR peak/shelf filters, firequalizer allows you to define an exact target response curve across the spectrum from 0 Hz to the Nyquist frequency (half your sample rate).

You define your custom curve using one of two parameters:

  1. gain_entry: Defines the curve by setting explicit frequency-gain pairs that FFmpeg interpolates between.
  2. gain: Defines the curve continuously using a mathematical formula evaluated as a function of frequency (f).

Method 1: Specifying Points with gain_entry

The gain_entry parameter is the most straightforward method for creating arbitrary equalizer curves. It takes a list of entries formatted as entry(frequency, gain_in_dB).

Basic Syntax

ffmpeg -i input.wav -af "firequalizer=gain_entry='entry(f1, g1); entry(f2, g2); ...'" output.wav

Example: Multi-Point Custom Curve

To apply a 6 dB boost at 80 Hz, a cut of 3 dB at 400 Hz, a 4 dB boost at 3 kHz, and a steep roll-off above 10 kHz:

ffmpeg -i input.wav -af "firequalizer=gain_entry='entry(0, 0); entry(80, 6); entry(400, -3); entry(3000, 4); entry(10000, 0); entry(14000, -24)'" output.wav

Note: Always anchor the curve at 0 Hz and at or near your maximum expected frequency to prevent undefined gains at the frequency boundaries.

Method 2: Mathematical Expressions with gain

The gain parameter lets you describe the target gain in dB as an algebraic formula using FFmpeg's expression evaluator. The variable f represents the frequency in Hertz.

Basic Syntax

ffmpeg -i input.wav -af "firequalizer=gain='EXPRESSION'" output.wav

Example: Conditional Band Shelf

To apply a 6 dB boost only between 1 kHz and 4 kHz, while keeping all other frequencies flat (0 dB):

ffmpeg -i input.wav -af "firequalizer=gain='if(between(f, 1000, 4000), 6, 0)'" output.wav

Example: Continuous Slope (Tilt EQ)

To apply a continuous tilt across the spectrum:

ffmpeg -i input.wav -af "firequalizer=gain='3 * log2(f / 1000)'" output.wav

Essential Configuration Parameters

To ensure the filter interprets your custom curve accurately and without audio artifacts, adjust the following optional settings:

1. Frequency and Gain Scaling (scale)

By default, FFmpeg interpolates frequencies linearly. Because human hearing is logarithmic, frequency points should usually be interpolated logarithmically by setting scale=loglin:

ffmpeg -i input.wav -af "firequalizer=scale=loglin:gain_entry='entry(20, 0); entry(100, 4); entry(1000, -2); entry(10000, 3)'" output.wav

Available scale options:

  • linlin: Linear frequency scale, linear gain scale.
  • linlog: Linear frequency scale, logarithmic gain scale.
  • loglin (Recommended): Logarithmic frequency scale, linear gain (dB) scale.
  • loglog: Logarithmic frequency scale, logarithmic gain scale.

2. Accuracy and Resolution (delay)

The delay parameter (measured in seconds) controls the length of the impulse response filter. A larger delay allows for steeper slopes and more precise low-frequency control, at the cost of higher latency and CPU usage:

ffmpeg -i input.wav -af "firequalizer=delay=0.05:gain_entry='...'" output.wav

The default delay is 0.05 seconds (50 ms), which is typically sufficient for most curves. Increase this up to 0.1 or higher if you need sharp cuts at sub-bass frequencies below 100 Hz.

3. Window Function (wfunc)

The window function shapes the transition of the FIR filter kernel. The default is hann, but you can change it to reduce ripple:

  • hann: Balanced roll-off and stop-band attenuation (default).
  • blackman: Provides greater stop-band rejection with slightly wider transition bands.
  • rectangular: Sharpest transitions, but introduces significant frequency rippling.
ffmpeg -i input.wav -af "firequalizer=wfunc=blackman:gain_entry='...'" output.wav

Complete Production Example

The following command applies a smooth, log-interpolated headphone compensation curve with high precision (delay=0.1) and high dynamic range windowing (wfunc=blackman):

ffmpeg -i source.flac -af "firequalizer=scale=loglin:delay=0.1:wfunc=blackman:gain_entry='entry(20, 3); entry(100, 1.5); entry(800, -1); entry(3500, 4); entry(7000, -2); entry(12000, 1); entry(20000, -4)'" -c:a flac corrected_output.flac