Approve the mix before mastering
Mastering should begin with a performance and mix you want to keep. Listen without stopping and mark lyric errors, abrupt transitions, distracting vocal textures, weak endings, and balance problems. Fix those in the arrangement or mix where possible. A limiter cannot restore a missing syllable, and extra brightness can make a metallic vocal more conspicuous.
Save the original download and a clearly identified premaster. The export settings guide covers obtaining and checking that source. Make sure the project is not clipping internally and that every intended instrument is audible before you evaluate overall loudness.
Do not confuse ordinary mastering with a generator’s similarly named feature. Suno’s Remaster documentation describes new variations of a clip. A conventional mastering chain processes your selected recording. Compare any generated variation as a new musical candidate, including the words and performance.
Harsh highs: build a small listening checklist
Choose reference music with a similar arrangement and intended listening context. A sparse acoustic song is a poor loudness target for a dense dance track, even when both sound good. Match reference playback volume to your mix before judging tonal balance.
Mark four passages in your own song: the opening, an exposed vocal or instrument, the densest chorus, and the ending. Add any known problem point. These suggested checkpoints make comparisons repeatable; they are not evidence that a particular AI model succeeds or fails on those passages.
Write a concrete aim for the master, such as “slightly firmer low end while keeping the chorus vocal clear.” Avoid vague instructions like “make it professional.” If the intended change cannot be described, it will be hard to decide whether another processor helps.
Overcompression: apply only processing with a clear purpose
Begin with modest tonal adjustment, listening across the marked passages after each change. If a cut fixes the chorus but hollows out the verse, consider automation or return to the mix. Whole-song processing is a blunt answer to a local problem.
Use compression only if the dynamic behavior needs it. Listen to whether the kick still has impact, the vocal still breathes, and quiet textures rise unnaturally between hits. An already dense source may need little additional compression. Reduce or bypass processing that wins only by being louder.
Use a limiter as a controlled final stage where needed, then compare against the premaster at equal perceived loudness. If attacks become smeared, low notes distort, or consonants turn brittle, reduce the drive. For persistent vocal trouble, return to metallic vocal repair instead of pushing the master harder.
Check loudness without treating it as an admission test
Measure integrated loudness and true peak with suitable meters, but interpret the numbers alongside listening. Integrated loudness summarizes a program; it does not describe every phrase. True peak estimates peaks that can occur between samples, making it a useful delivery check beyond a basic sample-peak meter.
Spotify’s guidance describes playback normalization and recommends a maximum true peak below −1 dBTP, with more headroom for louder masters. Its normal playback reference is −14 LUFS. That is playback guidance, not a rule that every recording must be mastered to exactly that number to be accepted.
Check the current delivery documentation of your distributor or client. Preserve dynamics that serve the song, and do not destroy them to hit a target borrowed from another service. If your tool creates a “Spotify” preset, verify what it actually changes rather than assuming the label establishes compliance.
Make a fair comparison and inspect the export
Line up premaster and candidate at the same start, then reduce the louder playback until the comparison feels even. Switch over the same passages with no other player enhancements changing. Ask whether the candidate improves balance and impact, and whether it introduces fatigue or distortion.
Listen in stereo, check mono compatibility, and try a second familiar system. Mono is useful for detecting disappearing elements; it does not mean every stereo difference is a fault. Review the whole track after short comparisons so the master still feels coherent from start to finish.
Export using the recipient’s requested format. Open that actual file and verify duration, channel count, start, tail, and measured peaks. Name it clearly and keep the premaster, project, and processing notes. Approval applies to this file, not merely to what played inside the project.
Complete release records separately
DistroKid permits AI music subject to its requirements, including rights and content rules. A pleasing master does not resolve those checks. Use the release-ready workflow for source records, accurate credits, and distributor requirements.
If you are considering automated processing, our Undetectr review explains why we recommend it for Suno release preparation. Evaluate any candidate by audible improvement and documented behavior; this checklist does not certify detector outcomes, watermark removal, or platform acceptance.