PRACTICAL GUIDE

How to check if your mix improved

A practical method to compare mix versions: matched loudness, blind listening, mono, clear questions and useful notes.

On this page8 sections

Compare the piano examples at matched loudness →

For a first try, load one of the piano examples. Brightness reduces B’s high frequencies with a filter: compare timbre and attacks with matching on. Loudness uses the same audio at different gains: matching should almost remove that difference. They demonstrate different things. Not hearing a relevant change is a valid result, not a failed task.

Three steps in the lab: Prepare loads A/B and sets the whole common range or an excerpt. Compare lets you play both and choose A, B or No relevant difference. My decision shows the files, passage and choice before exporting a report. Notes, listening options and blind trials are optional. Stop returns to the passage start; Pause keeps your position.

Start with a question you can hear

You finished a new mix and it seems better. Before replacing the previous version, write a specific question: “Is the chorus vocal clearer without losing the kick attack?” This names both the intended improvement and its possible cost.

The Delivery review and A/B comparison combines a passage, loudness matching, blind preference listening and notes. It does not grade your song. It helps you decide what to keep.

1. Prepare comparable files

Export A and B from the same point, at the same sample rate, without different export normalization. Use audio you own or have permission to use. Adjust each version’s start if one contains leading silence. Choose a short passage containing the problem: ten seconds of a chorus can be more useful than a whole song without a listening question.

Inputs must be mono or stereo, no longer than 10 minutes and no larger than 128 MB per file. Both limits apply. Compressed files use more memory when decoded. On a phone, a short copy of the passage can be easier to process. See preparing audio for analysis.

2. Remove the volume advantage

A small gain increase can make a mix seem clearer or more powerful. Turn on LUFS matching before deciding. In the lab’s original example, B is the same signal with more gain: matching should almost remove the difference.

The tool attenuates to a common level and allows peak headroom; it does not compress the files. Integrated loudness is measured across each complete file. If the new chorus changed relative to the rest of the song, global matching does not ensure equal loudness in that chorus. For a more controlled comparison, export the same excerpt from both versions and compare those files.

3. Listen for one change at a time

QuestionWhat to hearWhat to note
Is the vocal clearer?Consonants and phrase endings within the mix.Clarity versus mere presence. The vocal band filters all audio; it does not isolate the singer.
Do kick and bass coexist?Attacks and sustained notes in the same passage.Which instrument masks the other. The bass band focuses attention without deciding the balance.
Does the music still move?Attack, decay and changes between soft and loud sections.Whether compression solved the problem or flattened the performance.
Does it work in mono?Vocal, snare and elements that lose presence when channels are summed.Losses that affect the song, without treating narrower stereo as a fault by itself.

Listen to the complete mix first, then a band. Filters change both perception and loudness. Return to the full mix before keeping an adjustment. Keep playback volume moderate and take breaks.

4. Hide which version is new

Start a blind listen. X and Y are randomly assigned. Switch versions within the same passage, choose a preference and reveal the result. “No preference” is useful: the change may not matter, or you may need a different passage.

Repeat later if the decision matters. These preferences do not establish statistical significance or replace an ABX test. They help reduce the expectation created by seeing “final version” in a filename.

During a blind trial, passage and listening settings stay fixed. Listen to X and Y before choosing. To change settings, exit without saving and start a new trial. Revealed choices keep their original context. Up to 50 choices can be saved per file pair; replacing either file clears the history, while keeping your notes.

5. Read measurements in context

LUFS describes measured loudness; dBTP estimates peaks between samples. Neither is a quality score. A peak ceiling protects headroom without fixing a masked vocal. There is no LUFS number that makes every song professional.

HertzKit algorithms are based on ITU-R BS.1770; results are not certified measurements. Use the analyzer and mix review to investigate, and check the destination’s current requirements when delivering a master.

Write a note that prevents starting over

“At 00:35–00:48 the vocal is clearer in B. The chorus survives mono, but the kick lost attack. Keep the EQ and reduce bus compression.” This names a passage, a benefit, a cost and an action. It is more useful than “B has more air.”

Review your choice in My decision, then download the TXT report or the optional JSON notes before leaving. JSON contains notes, settings, measurements and preferences, without audio. Opening it restores notes; load the files again. Nothing is saved automatically. Date and number your exports, and keep a copy of the original files.

From listening to the approved WAV

Load the previous version in A and the delivery WAV in B. Review marked passages or record a full listen. Check the requirements, approve B and download the receipt. If you need to resume later, keep the draft JSON with both WAV files.

Open Delivery review → · Verify an approved WAV →