Setting Up Corrective EQ

What Corrective EQ Needs

As a standalone application, Corrective EQ sits between your source and your headphones. Audio leaves the app you are listening to, passes through a virtual audio device, enters Corrective EQ for processing, then returns to your headphones.

Four conditions have to be true for that path to work. If all four are true, Corrective EQ works. If one is false, it does not.

1. A virtual audio device is installed on your computer.
2. Your audio source’s output is set to that virtual device.
3. In Corrective EQ’s audio settings, the input is that virtual device and the output is your headphones.
4. Every device in the chain is running at the same sample rate.

Condition 1 — A Virtual Audio Device

A virtual audio device is a software audio path. It appears as an output on one side and an input on the other, so one application can send audio into it and another can receive that same audio back out.

You need it because the operating system does not normally let one application read another application’s output directly. The virtual device is the bridge between your player and Corrective EQ.

Use any virtual audio device you trust. It does not need to be ours. It only needs to run in stereo, support the sample rate you use, and appear inside Corrective EQ.

Condition 2 — Send Your Audio To It

Set the output of the app playing audio to the virtual audio device.

You can do that at the system level or inside an individual application. System output sends everything through the device. Per-app output routes only that app and leaves the rest of the system alone.

macOS — system output set to a virtual device

Windows — system output set to a virtual device

Ignore the exact device names. The selected output needs to be a virtual device, not your headphones or speakers.

Condition 3 — Tell Corrective EQ What To Use

Open Corrective EQ’s audio settings. Set two devices and make sure they are the right way round.

The input is the virtual audio device. The output is your headphones, or the interface they are connected to.

Reversing them is the most common mistake and it usually fails without any warning. Keep the direction simple: the virtual device goes in, your headphones go out.

Condition 4 — One Sample Rate Everywhere

Every device in the chain has its own sample rate setting, and they need to match: the virtual audio device, Corrective EQ, and your headphone output.

When they do not, the chain either resamples badly or stops working. The usual symptoms are clicks, crackles, stuttering, wrong pitch, wrong speed, or silence.

Pick one rate and use it everywhere you can set one. 44.1 kHz and 48 kHz are both safe. Most of these controls live in the operating system’s audio device settings.

macOS — every device carries its own format setting

Windows — the same setting, one device at a time

The point is not the example rate. It is the number of devices involved. Each one has its own format setting, and the ones in your signal path need to match.

Confirm Audio Is Arriving

Corrective EQ shows an input meter beside its input selection, in the same panel shown above. Play audio and watch it.

If the meter moves, the routing is correct. Audio is reaching the application. Any remaining problem is inside the app, not in the system routing.

If the meter does not move, audio never arrived. The fault is in one of the first three conditions, and changing processing settings will not fix it.

Check this before anything else. It tells you which half of the chain needs attention.

Import A Correction Curve

Corrective EQ does nothing until you import a correction file. It is not an equalizer you tune by ear. With no file loaded, audio passes through unchanged, and there are no factory presets.

If the application is running, the input meter is moving, and nothing sounds different, check this step first.

After importing, confirm two things: the curve appears in the preview, and the filter status says it is ready. The correction is not active until that render finishes.

What Counts As A Correction File

Four extensions are accepted — .csv, .txt, .eq, and .apo — but the content is what matters. Supported files fall into three formats:

• Frequency and decibel pairs. Two columns for one correction applied to both ears, or three columns for separate left and right correction. Header rows are fine, and at least two valid points are required.
• Parametric filter lists. A preamp line followed by peaking and shelving filters with frequency, gain, and Q. Disabled filters are skipped.
• Single-line graphic EQ strings. A list of frequency and gain pairs separated by semicolons, on one line.

If a curve does not cover the full audible range, its endpoint values are extended outward.

Where Correction Files Come From

Most correction files come from one of two workflows:

• Headphone correction. A headphone model is measured against a target response and the difference is saved as a file. You load the file for your exact model. It corrects the headphone, so every owner of that model can use the same file.
• Room correction. You measure your own speakers in your own room with a microphone and measurement software. The resulting file corrects that room, so it is specific to your setup and listening position.

The file describes the correction. Corrective EQ applies it; it does not create it.

What Correction Changes

Every headphone and every room has response peaks and dips. Peaks overstate part of the spectrum. Dips understate it. Over time, your brain starts to treat those deviations as normal.

A correction curve is the inverse of that response. Applied properly, it reduces the peaks and fills the dips so the tonal balance is closer to the source.

This is mainly a change in tone, not space or width. Listen for bands that stop jumping out or disappearing.

How To A/B It

Once the levels match:

1. Choose music you know well. Familiar material is easier to judge here than new material.
2. Toggle the correction on and off, or sweep the correction amount between none and full. Use repeated passes.
3. Judge one region at a time: bass, vocal body, presence, then top end.
4. Then leave it on for a few tracks without switching.

What You Are Listening For

Common signs that a useful correction is doing its job:

• A resonance settles down. Notes stop jumping forward in one narrow band.
• Upper-mid harshness reduces. Vocals, cymbals, or guitars stop pushing too hard.
• Masked detail becomes easier to hear. Low-level parts and reverb tails stop getting buried.
• The mix balance feels more even. Fewer elements sound consistently too loud or too recessed.

If Full Correction Is Too Much

You do not have to use 100%. The mix control blends between the uncorrected and corrected signals.

That is often useful with aggressive measurements. Large, narrow corrections can sound over-processed at full strength even when the source data is valid. Partial correction is common, especially for room correction.

Headroom is the other reason to back off. Curves that fill deep dips require boost, and boost can clip. If you hear distortion after import, reduce the output gain before blaming the file.

The Corrective EQ user manual covers the rendering choices and the latency and processing trade-offs behind them.

When It Is Not Working

When The Correction File Will Not Load

Known-Good Examples

Everything below is an example, not a requirement. The four conditions above are what matter. If a product name, version, or screenshot here no longer matches your machine, substitute the current equivalent and keep going.

• Guide scope — standalone Corrective EQ on macOS and Windows 11.
• Virtual audio devices on macOS — BlackHole from existential.audio/blackhole. ARIA-Audio-VAD, included with our ARIA Studio installer, also works.
• Virtual audio devices on Windows — VB-Audio Hi-Fi Cable from vb-audio.com/Cable.
• Headphone correction files — the AutoEQ project publishes measured corrections for a large number of headphone models in the parametric and graphic formats described above.
• Parametric and graphic formats — both come from Equalizer APO.
• Where device sample rates live — Audio MIDI Setup on macOS. Per-device properties in Sound settings on Windows.
• Screenshots — taken at 44.1 kHz. The device names in them are from our test machines and will not match yours.

Summary

Corrective EQ needs the same four setup conditions as every standalone product: a virtual audio device, your source sent to it, the correct input and output assignments, and one sample rate throughout. It also needs a correction file. Verify the routing with the input meter, verify the correction with the curve preview and ready status, then level-match before you compare anything.