How to master Suno tracks for release: a practical workflow

You have a Suno track at the “I'm happy with the song” stage. The arrangement works, the vocal sits right, and you want it on Spotify next week. What stands between that file and a release-ready master? More than most workflows admit. Suno mastering is usually framed as a loudness problem — pull the file into a DAW, slap on a limiter, push it to -14 LUFS. That misses the bigger issue: the file you exported still has decoder-stage artifacts baked into it, and no mastering plugin can remove them. This guide walks through the actual workflow end to end: refine first, master second.

What “mastering” actually does

Mastering is the final polish before distribution. Its job is integrated loudness in the right neighborhood for your genre, a true-peak ceiling that survives platform transcoding, and final tonal balance so tracks sit together on a playlist or an album. The widely-cited figures you'll see — -14 LUFS for Spotify, -16 LUFS for Apple Music Sound Check, -14 LUFS for YouTube and Tidal — are playback normalizationtargets, not mastering targets. Platforms turn loud masters down to those levels; they're a ceiling on what playback will sound like, not a value engineers aim for in the limiter.

What current commercial masters actually hit is louder. A 2024 survey of the Billboard Hot 100 averaged -8.3 LUFS integrated. Hip-hop and EDM commonly sit -6 to -8 LUFS; pop and rock -8 to -10. Acoustic, jazz, and classical material correctly stays much quieter (around -14 to -18 LUFS) because dynamic range is the point. Knowing where your genre lives — and why — is the first step. The practical reason engineers don't aim for -14: Apple Music, YouTube, and Tidal normalize only downward, so a quiet master plays back permanently quieter than peers on those platforms with no upward boost.

What mastering does not do is fix what the source signal is made of. A limiter pushes everything up. If the input already has vocoder warble in the breath of a vocal, a slightly-too-tight transient on the kick, or a periodized shimmer in the high end, the limiter will preserve and sometimes emphasize those characteristics. The output is louder. It is not cleaner. This is the single most common mistake when releasing AI-generated tracks.

Why a Suno WAV is not enough on its own

Suno's export tiers, as of 2026, look roughly like this: Free and Basic give you MP3 only, Pro adds 44.1 kHz / 16-bit WAV, and Premier delivers 48 kHz / 24-bit WAV. So if you are on Premier, you already have a high-resolution WAV. Why refine?

Because the WAV is a container. Suno generates audio with an autoregressive transformer that emits discrete audio tokens, which a neural codec decoder — a stack of transposed-convolution layers — then renders into a 48 kHz PCM waveform. The artifacts listeners associate with AI music (narrow stereo, slightly-too-tight transients, characteristic high-frequency shimmer or smear, vocoder warble on vocals) are introduced at that decoder stage, not at the file-export stage. A 2025 paper from Deezer titled “A Fourier Explanation of AI-music Artifacts” showed that transposed-convolution upsampling in neural codecs periodizes the spectrum architecturally — the artifacts are not a bug, they are a property of how the decoder is built. By the time Suno wraps the PCM in a 48 kHz / 24-bit WAV, the fingerprints are already in the audio.

Mastering plugins react to whatever they are fed. They cannot reach back inside the decoder and re-render the file. That is what refinement is for.

The two-step workflow: refine, then master

1. Refine the Suno export. Upload your MP3 or WAV (up to 8 minutes per file). Pick a preset.

Enhanced uses a trained audio model and GPU processing to synthesize high-frequency detail. This is a plausible reconstruction, not recovery of the exact lost information: texture and timbre can change. It costs one credit and is the shortest processing path, although end-to-end time varies with track length, queue demand, and worker startup. It is the recommended starting point; compare the result with your original before mastering. Neither refinement preset should be chosen for, or marketed around, AI-detector behavior.

Flow Matching uses compute-intensive generative reconstruction, rebuilding detail through repeated refinement. Normal and High use different processing depths and require more GPU processing than Enhanced. Texture and transients can change, and more processing does not always mean a better result. It is temporarily unavailable while output quality is reviewed. Use Enhanced for new work. The comparison remains in Enhanced vs Flow Matching, and current availability is documented in the help page presets section. You can hear a real refinement before/after on the homepage.

Enhanced outputs 24-bit WAV at 44.1 kHz by default, and Pro and Studio unlock 48 kHz. The Flow Matching-only 32-bit float option is temporarily unavailable.

2. Master — in-app, or in a DAW. The Refiner can now do this step for you. Flip on Finalize(Hobby, Pro, and Studio; no extra credits) and the refined file runs through an automated mastering chain: corrective EQ to tame the 2–5 kHz harshness Suno tends to push, gentle glue compression, valve-triode and transformer harmonic saturation, stereo polish with a mono-safe low end, and a look-ahead true-peak limiter set to a ceiling of -1 dBTP — at a loudness target that defaults to -9 LUFS integrated, the middle of the competitive band for pop and rock. You get both files back, mastered and pre-master, so you can A/B against your own chain. There's also a standalone Finalize preset (1 credit) that masters a track that's already mixed, refined or not.

Finalize is tunable rather than fixed. Strength comes in Gentle, Medium, and Strong; the loudness target defaults to -9 LUFS integrated with -7, -14, and -16 options for different release contexts; harmonics run Off, Standard, or Hot. Optional reference matching lets you upload a track whose tone you like, or pick one of the curated factory presets. The free 10-second reference preview auditions only the selected reference's tonal shape; with Reference Off it plays the original excerpt. Strength, loudness, and harmonics are applied when the final run starts and are not rendered in that preview.

For full control, master the pre-master in a DAW instead. A working chain looks like this: corrective EQ to tame anything harsh in the 2–5 kHz range, optional gentle compression for glue at roughly 1.5:1 with slow attack and slow release and a decibel or two of gain reduction, then a true-peak limiter set to a ceiling of -1 dBTP (or -2 dBTP if you're mastering above -14 LUFS integrated, to leave Spotify's Ogg/AAC encoder enough headroom). Push the limiter until integrated LUFS lands in your genre's competitive band — not the playback normalization figure, the actual mastering band a commercial track in your style hits. Reference a real chart track and match by ear. Free DAWs like Reaper, Tracktion Waveform Free, or Cakewalk all run a full mastering chain.

One technique worth knowing: instead of pushing a brickwall limiter harder to get loud, raise the average from below. Upward compression lifts material under a threshold while leaving peaks alone. Parallel (“New York”) compression blends a heavily compressed copy under the dry signal for density without smashing transients. Multiband upward processors— iZotope Ozone's Upward Compress module, RX's Loudness Control, FabFilter Pro-L 2 in modern modes — handle this per-band so you don't muddy the low end while lifting vocal density. Integrated LUFS rises, transients survive, and the track does not sound smashed. The peak-to-loudness gap (sometimes called PSR or crest factor) is what perceived dynamics is about; staying above PSR ≈ 8 during the loudest passage keeps the master from clamping audibly. Judge how far to push any refined source by its own dynamics rather than by which refinement preset produced it.

Format and sample rate for the final master

Most distributors accept 16-bit or 24-bit WAV at 44.1 kHz. Some accept FLAC. None ask for MP3 as a master, and any distributor that does should be treated with suspicion. Bouncing your final loudness pass at 24-bit, 44.1 kHz WAV puts you inside the spec for Spotify, Apple Music, Amazon Music, Tidal, YouTube Music, and the major aggregators. The full distribution checklist — artwork sizes, ISRC codes, metadata fields, and the platform-specific AI disclosure rules — lives in Preparing AI music for distribution.

Common pitfalls

Treating a Suno WAV as a finished master. A WAV is a high-fidelity container around a neural codec output. The decoder fingerprints are inside the PCM. Mastering plugins polish symptoms; refinement addresses the layer where the artifacts were introduced.

Limiting on top of an older 32-bit float Flow Matching bounce without checking peaks. That format was intended for re-entry into a DAW so that intersample peaks retained headroom during downstream gain; Flow Matching and its 32-bit option are currently unavailable. If you are sending a file straight to a distributor, render the master to 24-bit and confirm the true-peak ceiling sat at -1 dBTP before bounce.

Over-compressing already-tight transients.Suno's internal output already has tight, controlled transients. Adding heavy bus compression on top flattens what little punch is left. Glue compression should barely move the needle.

Trusting plugins that claim to “de-AI” a track. Mastering-chain plugins operate on the final mix and cannot reach the decoder stage where the artifacts were introduced. Anything that promises to undo AI character with a single insert is, at best, doing EQ and stereo widening with a marketing label on top.

Pushing for streaming-loud levels on an unrefined Suno mix. The decoder artifacts can already constrain transient detail; a brickwall limiter on top may emphasize the problem and create pumping. Refine first with Enhanced as the default, compare alternatives only if needed, and then set loudness by listening to that specific mix rather than assuming one preset can safely take a harder master.

Try the complete workflow

The Refiner now covers the whole workflow: refinement handles the part Suno's decoder bakes in — the part mastering plugins cannot reach — and Finalize handles loudness, ceiling, and final polish. If you'd rather have a mastering engineer's hand on it, the pre-master file is a cleaner starting point than a raw Suno export. The Free tier gives you two credits a month to try an Enhanced refinement on a real Suno track; Finalize is available on Hobby, Pro, and Studio — pricing for higher volumes is on the pricing page. Join the waitlist from any page.