What sample rate and bit depth should I deliver?

16-bit/44.1 kHz is the floor most distributors accept as of writing. 24-bit at 44.1 kHz or 48 kHz is the more common preference and the format The Refiner exports by default. Pro and Studio plans can render at 48 kHz, and Flow Matching on those plans can export 32-bit float for further DAW processing. Check your distributor's current documentation.

What loudness should I master to?

Genre-competitive, not -14 LUFS. The -14 figure is Spotify's playback normalization target — where they bring loud masters down to, not a mastering target. The 2024 Billboard Hot 100 averaged around -8 LUFS integrated; hip-hop and EDM commonly sit -6 to -8, pop and rock -8 to -10. Acoustic, jazz, and classical material correctly stays around -14 to -18 LUFS. Pick the band your genre lives in, use a true-peak ceiling near -1 dBTP, and let the platforms normalize on their end. The full walkthrough — chain, upward-processing techniques, ceilings — is in the Suno mastering guide.

Do I have to disclose that my track was made with AI?

Check your distributor's current AI-music disclosure policy before submission — most major platforms have introduced or are rolling out AI-disclosure flags, and the rules are still evolving. This is not legal advice. The low-risk default is to mark the track as AI-generated or AI-assisted where the form asks, and to keep your own records of which tools you used.

Does The Refiner master tracks for release?

No. The Refiner rebuilds audio fidelity — it addresses the decoder-stage fingerprints from the generator. Loudness mastering, limiting, LUFS targeting, and submission to a distributor are separate steps you handle in a DAW or with a mastering tool after refinement. We're explicit about that scope so you can plan the rest of the workflow.

How to distribute AI music: a pre-release checklist for Suno tracks

Published May 25, 2026

You picked an aggregator. You did not pick what's inside the WAV you upload. That's the part most release-prep articles skip. If you want to distribute AI music to Spotify, Apple Music, Bandcamp, or SoundCloud, the submission form is the easy step; the audible "AI sound" survives the upload, the transcode, and the listener's earbuds. Distribution prep is more than a file-format conversion, and the old advice of "just convert your mp3 to a WAV" misses the actual problem.

This guide walks through what actually happens between a Suno generation and a release-ready submission: where the AI artifacts come from, which step fixes which problem, and what The Refiner does and deliberately does not do. Four steps, in order: refine the fidelity, master in a DAW, confirm format requirements, handle AI disclosure.

Why a Suno WAV isn't automatically release-ready

As of writing, Suno's free and Basic tiers download as MP3 only. Suno's Pro tier exports WAV at 44.1 kHz / 16-bit, and the Premier tier exports at 48 kHz / 24-bit. So a substantial slice of Suno's paying users already have a WAV in hand and don't need a format-conversion step at all. That sounds like the end of the story — lossless container, distributor-friendly spec, done. It isn't. The artifacts that make AI music sound subtly off were stamped into the PCM samples before any file format wrapped them.

AI music generators work in roughly this order: a transformer language model emits discrete audio tokens, and a neural codec decoder — a stack of transposed-convolution upsampling layers — turns those tokens back into 48 kHz PCM. That decoder stage leaves a structural fingerprint in the spectrum. A 2025 Deezer paper titled "A Fourier Explanation of AI-music Artifacts" showed that transposed-convolution upsampling architecturally periodizes the spectrum, producing model-specific patterns that listeners hear as narrow stereo image, slightly-too-tight transients, high-frequency shimmer or smear, and vocoder warble on sustained vocals.

Because those artifacts are in the PCM, a lossless WAV preserves them losslessly. An MP3 layers psychoacoustic loss on top but doesn't remove the underlying decoder peaks. So the Pro/Premier WAV download is a better starting point than the free-tier MP3, but neither one is a finished release. The decoder fingerprint travels with the audio either way. We unpack that fingerprint layer in more detail in why Suno tracks sound compressed.

Step 1: refine the fidelity

Refinement is the only step in this workflow that addresses the decoder fingerprint itself. Loudness mastering, format conversion, and AI disclosure don't touch it — they happen further downstream. Drop the Suno file (WAV or MP3) into The Refiner and pick a preset. Enhanced is fast and runs in seconds — it polishes the top end and tightens the mix without re-synthesizing the audio, which makes it the right call when you mostly trust the source and just want a cleaner version of it. Flow Matching is slower and more aggressive: it reconstructs the track from scratch, which is what washes out vocoder-style fingerprints most thoroughly. If the generator output still sounds psychoacoustically processed after Enhanced — soft transients, blurred cymbals, washed reverb tails — run the source through Flow Matching for a deeper pass.

Both presets export WAV at 44.1 kHz, 24-bit by default. Pro and Studio plans unlock 48 kHz, and Flow Matching on those plans can render 32-bit float so the file can go straight into a DAW without re-quantization in the middle of the chain. The refined WAV is what you carry into step 2 — not the raw Suno download, even if your tier gave you a WAV. The container is the same; the contents are not.

Step 2: master in a DAW

Mastering is where loudness, true-peak control, and final tonal balance get handled. The Refiner does not do this. We don't run a limiter, we don't target a LUFS number, and we don't apply platform-specific mastering profiles. That is a deliberate scope decision: those choices belong to you (or a mastering engineer, or a mastering plugin) at the end of the chain, after fidelity is sorted.

Streaming platforms normalize on playback — and the widely-cited figures (-14 LUFS for Spotify, YouTube, and Tidal; -16 LUFS for Apple Music Sound Check) are playback normalization targets, not mastering targets. Platforms turn loud masters down to those levels; they don't say "this is what to aim for in the limiter." Commercial chart masters run louder than that.

What current commercial masters actually hit (2024–2026):

Hip-hop, EDM: roughly -6 to -8 LUFS integrated.
Pop, rock: roughly -8 to -10 LUFS integrated.
Billboard Hot 100 average (2024): around -8 LUFS.
Acoustic, jazz, classical: -14 to -18 LUFS — and rightly so, because dynamic range is the point.

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. Spotify's default Normal mode is the exception (it does boost quiet tracks up, within limited headroom). Mastering at genre-competitive loudness keeps a track on equal footing across playback contexts. Bandcamp and SoundCloud don't aggressively normalize at all — deliver the master at the level it's meant to play at.

Aim for the right neighborhood with a true-peak ceiling around -1 dBTP (or -2 dBTP if you're mastering above -14 LUFS, to leave Spotify's Ogg/AAC encoder enough headroom). The longer walkthrough — chain, upward-processing techniques, PSR / crest factor — lives in how to master Suno tracks for release.

Step 3: confirm format requirements

Most major aggregators — DistroKid, TuneCore, CD Baby, AWAL, Amuse — accept WAV. The common spec is 16-bit / 44.1 kHz minimum, with 24-bit at 44.1 or 48 kHz preferred. Some accept FLAC or AIFF as well; none of the major paid distributors want a lossy MP3 as a release master. Specs shift, so check your distributor's current documentation before you submit rather than relying on a one-line summary from anyone, including us.

The Refiner's default 24-bit / 44.1 kHz WAV lands inside that preferred range and matches what most aggregators recommend. If a distributor's documentation specifically asks for 48 kHz (some video-adjacent or sync-focused tiers do), Pro and Studio plans render at 48 kHz. Stereo only — surround and Atmos releases have separate workflows that are outside the scope of this guide. If you're delivering to a platform with a specific bit-depth or sample-rate requirement that doesn't match your DAW session, do the sample-rate conversion once at the end of your chain, not in the middle of it.

Step 4: handle AI disclosure

Major platforms have introduced — or are in the process of rolling out — AI-music disclosure flags and metadata fields. The exact labels and rules differ between distributors and stores, and they are still evolving. We're not going to quote specific store policies that will be out of date by the time you read this, and we're not giving legal advice.

The honest default: check your distributor's current AI-music disclosure policy before submission, flag the track as AI-generated or AI-assisted where the form asks, and keep your own records of which tools were used at which step — the generator, the refinement preset, the mastering chain, any stems work. That position ages well across policy changes and store-side audits, and it's also cheaper than having a release pulled after the fact.

Beyond disclosure, every distributor asks for the same basic metadata: track title, artist name, ISRC (your aggregator can mint one for you), release date, genre, language of the lyrics, explicit flag, cover art at the resolution they specify, and writer or publisher credits where applicable. Have those ready before you start the submission flow so the file you uploaded earlier in the session doesn't time out.

A note on Udio

If your source is Udio rather than Suno, the upstream of this workflow has shifted recently. Downloads were disabled across Udio in late October 2025 as part of an industry settlement, and a new partner-built platform is planned for the first half of 2026 without raw audio downloads. The practical implication: a substantial Udio audience currently can't submit new material to a distributor without having downloaded the file before the change. The rest of this workflow still applies to anything you saved earlier.

Pre-submission checklist

Run this before you hit submit. It's the same list we'd use on our own track.

Source has been refined. The decoder-fingerprint layer has been addressed with Enhanced or Flow Matching.
File format: WAV (or FLAC if your distributor prefers it), 24-bit, 44.1 or 48 kHz, stereo.
Loudness has been handled in a DAW or with a mastering tool. True peak around -1 dBTP. Integrated loudness in the right neighborhood for your target platforms.
No clipping. Look at the file in a DAW, not just your ears.
Metadata complete: title, artist, ISRC, release date, genre, language, explicit flag, cover art at the right resolution, writer credits.
AI disclosure handled per your distributor's current policy. Records kept of which tools were used.
Final master listened to end-to-end on at least two systems, one of them not headphones.

Where The Refiner fits

Distribution is multi-step, and we only solve one of those steps well: the fidelity step. We rebuild the audio so the decoder fingerprint stops shouting "this came out of a generator." Loudness mastering, metadata entry, AI disclosure, and the upload itself are separate tools and separate decisions, and we'd rather say so than oversell.

Free tier covers 2 refinements a month, which is enough to refine a single track twice (once with Enhanced, once with Flow Matching) and compare them before you commit to a release version. Plan details and credit packs are on the pricing page. The homepage at The Refiner has side-by-side audio so you can hear what each preset does before you upload anything. Start free, refine the track, then carry the WAV into your DAW for the mastering and submission steps.

Frequently asked questions

Isn't a WAV from Suno already release-ready?: Not quite. WAV is lossless, so the file format is fine for a distributor, but the audio inside that WAV still carries decoder-stage fingerprints from the generator — narrow stereo, slightly-too-tight transients, high-frequency shimmer, vocoder warble on vocals. WAV preserves those artifacts losslessly. If you care about the AI-generated sound, the file format is the easy part; the audio content is what you treat next.
What sample rate and bit depth should I deliver?: 16-bit/44.1 kHz is the floor most distributors accept as of writing. 24-bit at 44.1 kHz or 48 kHz is the more common preference and the format The Refiner exports by default. Pro and Studio plans can render at 48 kHz, and Flow Matching on those plans can export 32-bit float for further DAW processing. Check your distributor's current documentation.
What loudness should I master to?: Genre-competitive, not -14 LUFS. The -14 figure is Spotify's playback normalization target — where they bring loud masters down to, not a mastering target. The 2024 Billboard Hot 100 averaged around -8 LUFS integrated; hip-hop and EDM commonly sit -6 to -8, pop and rock -8 to -10. Acoustic, jazz, and classical material correctly stays around -14 to -18 LUFS. Pick the band your genre lives in, use a true-peak ceiling near -1 dBTP, and let the platforms normalize on their end. The full walkthrough — chain, upward-processing techniques, ceilings — is in the Suno mastering guide.
Do I have to disclose that my track was made with AI?: Check your distributor's current AI-music disclosure policy before submission — most major platforms have introduced or are rolling out AI-disclosure flags, and the rules are still evolving. This is not legal advice. The low-risk default is to mark the track as AI-generated or AI-assisted where the form asks, and to keep your own records of which tools you used.
Does The Refiner master tracks for release?: No. The Refiner rebuilds audio fidelity — it addresses the decoder-stage fingerprints from the generator. Loudness mastering, limiting, LUFS targeting, and submission to a distributor are separate steps you handle in a DAW or with a mastering tool after refinement. We're explicit about that scope so you can plan the rest of the workflow.