AI Music Mastering: Compare the Sound, Not Just the Loudness

The mastered version is louder, so it wins the first listen. Turn it down and the vocal may be harsher, the snare flatter, and the ending clipped. A useful comparison controls loudness before deciding whether the processing helped.

Download the mastering listening log · Create a cleaner music source

Editorial guide checked September 13, 2026. Examples and worksheets are original planning material; the hero is an AI-generated illustration, not a model benchmark.

Decide whether the source is ready for mastering

Mastering processes a finished mix for its intended delivery. It can change tonal balance, dynamics, and overall level, but it is not a substitute for choosing a good composition or repairing the arrangement. Before comparing presets, listen to the unprocessed source all the way through and identify defects that belong earlier in the workflow.

For this original exercise, imagine a ninety-second instrumental with a quiet electric-piano opening, a fuller middle section, and a final chord that decays into silence. The musical goal is warmth and a clear pulse without losing the quiet opening. There is no measured audio benchmark attached to this example; the listening log is a method to apply to your own file.

If the middle contains a missing beat, the piano changes character abruptly, or the ending stops before the chord finishes, address those problems first. Making them louder or brighter does not make them correct. Keep a source-defect list separate from the mastering comparison so the tool is not asked to solve the wrong problem.

Keep the best available source

Export or download the highest-quality source actually available from the production workflow. Keep an untouched copy. Converting a compressed file into WAV does not recover information discarded earlier; it simply changes how the next version is stored. Avoid repeatedly encoding intermediate copies when an uncompressed or lossless source is available.

Check that the source begins and ends where intended and that its loudest section is not already audibly distorted. If you control the mix, remove accidental clipping and reconsider excessive bus processing before mastering. Lowering the level of a clipped recording does not restore the original peaks.

BandLab’s mastering FAQ explicitly warns that mastering cannot repair a problem mix. Its preparation advice belongs to that service’s workflow; use the instructions of your selected tool rather than treating every suggested input level as a universal law.

Choose two treatments with a clear purpose

Start with the original and two processed versions. One might use a restrained treatment and another a stronger tonal or dynamic change. Record the preset and settings. Naming versions “A gentle” and “B bright” is more useful than collecting a dozen files whose differences you cannot reconstruct.

The current BandLab mastering instructions describe upload, preset selection, input-gain adjustment, and download, with some controls depending on membership. Check the controls in your account. A preset name expresses a suggested character; it does not establish that the result is correct for your track.

Use the same source for both treatments. If you compare one master made from a revised mix with another made from an older mix, you cannot isolate the effect of mastering. If the source changes, restart the comparison and keep the previous results clearly labeled as an earlier round.

Match listening level before choosing a winner

Place the original and processed versions on separate tracks in an audio editor, aligned to the same start. Use a loudness meter and playback gain to bring them to a similar perceived level over the same material. Do this for the comparison; do not overwrite the deliverable simply to make the audition convenient.

Peak matching alone is not enough. Two files can reach the same highest peak while one sounds much louder overall because its dynamics are more compressed. Compare the same passage at a comfortable monitoring level and switch between versions without changing the speaker volume. A level-matched audition makes changes in tone and transient shape easier to hear.

If you do not have a meter, a careful ear-based level match is still more informative than accepting the louder file immediately. Be honest about the limitation: it is a listening comparison, not a measured loudness report. Do not attach precise LUFS or true-peak claims to a file you have not measured.

Use three fixed listening windows

For the ninety-second example, mark the first ten seconds, a ten-second window in the fullest section, and the final ten seconds. Always compare those same passages. Moving to a different musical moment while switching versions makes the result harder to interpret.

Write observations without using preset names as explanations. “The bass masks the left-hand piano notes” is more useful than “Fire is too much.” “The final decay sounds smoother at matched level” supports a decision you can revisit. If neither treatment improves the source, keeping the original is a valid result.

Check the destination without chasing one magic number

Spotify’s normalization guidance describes a default playback target of minus fourteen LUFS and explains that normalization changes playback gain rather than rewriting the uploaded file. It also notes different listener settings and playback environments. That is context for Spotify delivery, not a universal mastering target for every use.

Follow the current destination guidance for file format, peaks, loudness, and metadata. A music release, a podcast, and a video soundtrack may have different requirements. If the track will sit beneath dialogue, judge it in that mix as well; a master that works alone can compete with the speaker.

Measure the exported file, not just the live project. Encoding or export settings can affect the deliverable. Reopen it, inspect its beginning and ending, and compare the relevant readings with the destination requirements. Keep those measurements attached to the exact filename so a later revision does not inherit an old approval.

Return upstream when the problem belongs upstream

If the arrangement is too dense, create a sparser source or adjust the mix. QuestStudio Music Lab can explore a revised musical brief, but the current workflow does not provide mastering controls. Use it for source creation, then perform the measured finishing and export in an audio tool.

If you regenerate, do not assume the melody or arrangement will remain identical. Review the new source as a new take. When exact continuity matters, work from editable tracks or a supported source-editing workflow instead of using a general text prompt as though it were a surgical mix adjustment.

Approve the master with a short listening record

After a break, listen to the chosen version on a second ordinary playback system at a comfortable level. Confirm that the musical priorities still hold: warm piano, a readable pulse, and an intact ending. Recheck the level-matched comparison if a brighter or louder presentation initially influenced the decision.

Save the untouched mix, selected master, settings, measurements, and listening notes together. The worksheet records why the version was accepted and what remains uncertain. A useful master is one that improves the track for its actual destination while preserving the musical decisions that made the source worth finishing.