Copy the balance of one recording onto another. Attach a reference, and the difference between its long-term spectrum and the loaded file’s is applied as a curve: a dull mix against a bright master gets its top lifted and its low mids pulled down by however much separates the two.
Attach a reference
The reference is read for its first 20 seconds at the loaded file’s sample rate, so start it inside the material. Its spectrum is measured in the same 96 bands and with the same window as the processing, which is what lets the two be subtracted band for band.
The source side is the loaded file, measured from start to end when it loads. To match a section instead, select it on the waveform and press the measure button: the curve is then computed against that section and applied to the whole file.
Amount, smoothing and limit
Amount scales the whole curve. 100 is the full difference; 50 is halfway, which is where most matches sound like the same mix with better balance rather than a different one.
Smoothing averages the curve across neighbouring bands with a triangular kernel. Low values track the reference’s own notes and resonances, which rarely belong on another recording. Around 40 the curve is a handful of broad moves. At 100 it is a tilt.
Limit caps every band at that many decibels of boost or cut. 12 dB is generous. A match that wants more than that is comparing two things that should not be matched: a voice against a full mix, or a file with no content in a band the reference fills.
In a chain
Inside the audio workflow and on a track in Oxyd the source is not measured first. The step averages what passes through it with a six-second memory and applies the correction gradually over the first three seconds. On a long file the curve ends up close to the fixed one. On a short one, measure the file on the tool page instead and let the chain carry the result.
What it cannot do
A spectrum match cannot copy compression, width, reverb or arrangement. Two mixes with identical long-term spectra still sound different. What it does fix is the balance decisions that depend on monitoring: a mix made on small speakers that has too much low mid, or a voice recorded a foot too close to the microphone.