How to import your own headphone measurement

Why you would import anything

Attune ships with measurements we made ourselves, on our own fixture. That list is deliberately short and grows one headphone at a time, because every entry is a real measurement rather than a number copied from someone else. If you own something we have not measured yet, importing is how you use the app today — and it is free, permanently. It is not a Pro feature and never was.

Where to get a file

Two free sources cover most of what exists:

• squig.link — a network of independent measurement databases. Find the site that covers your brand, search the model, and download its raw data.
• AutoEQ — a large open repository on GitHub with thousands of models, each in a folder containing a raw response file.

Some manufacturers publish their own measurements too. Those are fine to use, with the same caveat as everything else here: it came off their rig, not ours.

Which file to pick — the mistake worth avoiding

Most databases offer the same headphone two ways: a raw curve and a compensated one. Compensated means a target curve has already been subtracted, so the graph shows how far the headphone sits from that target rather than what it actually measured.

Attune wants raw. It applies your target itself. Give it a curve that already has a target baked in and the target gets counted twice — and the result of that is the genuinely dangerous part: it does not look broken. You get a tuning that seems reasonable, sounds off in a way you cannot name, and gives you no hint why.

How to tell them apart:

• A raw curve wanders. It has a rising region in the upper mids, real peaks and dips, and it does not sit near a straight line.
• A compensated curve hugs zero. If the whole thing hovers around a flat line with small wiggles, or the file or download button says compensated, Harman, diffuse field, DF, ordelta, it is not the one you want.

What accuracy to expect

Attune flags imported curves on screen as approximate, and that warning is doing real work. Voice-matching works by subtracting one measurement from another; the fixture’s own signature cancels outonly when both curves came off the same fixture. An imported curve came off someone else’s, so the difference between the two rigs stays in the answer.

In practice that means:

• Bass and midrange tonal balance — trust it. Rigs mostly agree here.
• Fine treble detail — do not. Above roughly 8 kHz, rigs disagree by several decibels, and no amount of careful listening will fix data that started out different.
• Coupler family matters. Curves from 711-style couplers transfer to our fixture better than curves from a 5128, because ours shares the same coupler standard.

And it is not your unit

Even a perfect measurement of a headphone is not a measurement of your headphone. Sample variation between two units of the same model is real, and pad wear moves the response further. An import is a good approximation of the model, not a portrait of the one on your head.

That is the honest ceiling on imported data, and it is why we offer amail-in measurement service — your actual unit, on our fixture, cancelling cleanly against our targets. Most people will never need it. If you are chasing the last decibel, it is the only way to get there.

The honest bottom line

Grab the raw file, import it, and enjoy a good tuning in the bass and mids. Ignore fine treble differences, remember it is the model rather than your unit, and do not let a compensated file in the door. Those three habits are the difference between an import that helps and one that quietly misleads you.

Common questions