Guide · Loudness
AI Music Loudness: LUFS Targets for Spotify & Apple Music (2026)
Updated June 2026 · 9 min read
Suno and Udio tracks arrive unmastered — often quiet, thin, and inconsistent section to section. Because streaming platforms normalise every track to their own reference, an unmastered AI song plays back weak next to commercial releases. This guide covers what LUFS and true peak mean, the per-platform targets, and how to hit them without pumping or clipping.
#What LUFS and True Peak Actually Mean
Two numbers decide how loud your track plays on streaming services:
- Integrated LUFS — Loudness Units relative to Full Scale, measured across the whole track. This is the perceived average loudness the platform reads to decide how much to turn your track up or down. Lower (more negative) means quieter.
- True peak (dBTP) — the highest level the waveform reaches, measured between samples so it catches peaks that a normal sample-peak meter misses. Lossy encoders (AAC, Ogg) can push true peaks higher during conversion, so leaving headroom below 0 dBTP prevents clipping on playback.
Streaming platforms apply loudness normalisation: they measure your integrated LUFS and adjust playback so every track sits near one reference level. Most major services turn loud tracks down but do not turn quiet tracks up — which is exactly why an unmastered AI export sounds small next to commercial releases.
#Why AI-Generated Tracks Sound Quiet, Thin, or Inconsistent
Generators like Suno and Udio render audio, not masters. Three things commonly go wrong:
- Conservative output level — the raw file often sits well below streaming reference, so after normalisation it plays back quieter and with less presence than mastered tracks around it.
- Thin or unfocused low end— bass weight and stereo balance are often uneven, which reads as "thin" once a track is sitting in a loud playlist.
- Section-to-section inconsistency — intros, drops and outros can vary in loudness within a single track, because the model never balanced them as a finished release.
Mastering fixes all three: it raises the integrated loudness toward the platform reference, controls the low end, and evens out the dynamics so the track holds its place in a queue.
#Per-Platform Loudness Targets (June 2026)
We're not affiliated with any streaming service. The values below reflect public documentation and widely corroborated reports as of June 2026, and platforms change these without notice — tap a name to confirm the current rule before you release. Most services cluster around −14 LUFS; Apple Music is the notable outlier at about −16 LUFS.
| Platform | Integrated LUFS | True peak | Notes (June 2026) |
|---|---|---|---|
| Spotify | −14 LUFS | −1 dBTP (−2 if louder) | Only turns louder tracks down; quieter tracks are not boosted. Official artist guidance. |
| Apple Music | −16 LUFS | −1 dBTP | Sound Check reference is lower than the others. Down-only; quieter tracks stay quiet. |
| YouTube / YT Music | ≈ −14 LUFS | −1 dBTP | Normalises loud uploads down only. Not officially published as a fixed number. |
| Tidal | −14 LUFS | −1 dBTP | Album-level normalisation preserves track-to-track relationships within a release. |
| Amazon Music | −14 LUFS | −2 dBTP | Strictest true-peak ceiling of the group — master to −2 dBTP to be safe everywhere. |
| SoundCloud | ≈ −14 LUFS | −1 dBTP | Behaviour is community-tested, not officially documented. Treat as approximate. |
Linked pages are each service's official help/policy site. Numbers are approximate where the platform does not publish a fixed value (YouTube, SoundCloud). Confirm the current rule on the exact platform before release.
#One Master for Everywhere — or One per Platform?
You almost never need a separate master per platform. Because each service normalises down to its own reference, a single master at about −14 LUFS integrated with −1 dBTP(or −2 dBTP to satisfy Amazon's stricter ceiling) travels cleanly across Spotify, YouTube, Tidal and Amazon, and is simply turned down a touch further on Apple Music's lower reference. The thing that matters is not chasing a different LUFS number for each logo — it is delivering one master with controlled true peak and natural dynamics.
#How to Hit the Target Without Pumping or Clipping
Reaching −14 LUFS is easy; reaching it cleanly is the craft. Common pitfalls and how to avoid them:
Use a genre-appropriate master, not a one-size limiter
A dense EDM track and a sparse acoustic ballad need different handling to land at the same integrated loudness. Pushing both through one aggressive limiter is what causes pumping (audible level breathing) and squashed transients. Genre-aware presets apply tone shaping and limiting suited to the material before the final loudness stage.
Leave true-peak headroom
Limit to a true-peak ceiling, not a sample-peak ceiling. A −1 dBTP ceiling (or −2 dBTP for Amazon and louder masters) keeps inter-sample peaks from clipping after lossy encoding. Aiming for 0 dBFS sample-peak is a classic cause of distortion that only appears after upload.
Master loudness, then measure
Set the integrated loudness as a deliberate target and verify it with a meter — don't eyeball the waveform. Over-compressing to chase a number that the normaliser will undo only sacrifices dynamics. If a track sounds great at −14 LUFS, there is nothing to gain by forcing it to −9.
#Common Loudness Mistakes With AI Tracks
| Mistake | What happens |
|---|---|
| Uploading the raw AI export unmastered | Plays quiet and flat next to mastered tracks after normalisation. |
| Pushing far past −14 LUFS | Normaliser turns it back down; you lose dynamics for no loudness gain. |
| Limiting to 0 dBFS sample peak | Inter-sample / true-peak clipping appears after lossy encoding. |
| One aggressive limiter for every genre | Pumping, squashed transients, and a fatiguing master. |
| Ignoring Amazon's −2 dBTP ceiling | A −1 dBTP master can clip on Amazon's stricter requirement. |
| Mastering each section by ear separately | Inconsistent loudness within the same track in a playlist. |
#One-Click Loudness-Correct Mastering
Anti-AI Master masters your AI track to a streaming-ready integrated loudness with a controlled true peak, using genre-appropriate presets so the result lands at target without pumping or clipping. The same studio also runs separate Anti-AI processing to lower the AI-detection confidence score — two independent jobs in one upload. For platform-specific release rules and the disclosure-vs-detection picture, see our AI music distribution guide. For generator-specific tone and level handling, see Suno mastering and Udio mastering.
#Frequently Asked Questions
What LUFS should I master my Suno or Udio track to?
A −14 LUFS integrated / −1 dBTP true-peak master is a safe single target for most major streaming platforms (Spotify, YouTube, Tidal and Amazon all normalised to about −14 LUFS as of June 2026; Apple Music uses about −16 LUFS). You don't need a separate master per platform — each service turns louder tracks down to its own reference, so one well-controlled master travels well. Always confirm the current rule on each platform before release.
Why do AI-generated tracks come out quiet or thin?
Most generators (Suno, Udio) export at a conservative level with limited low-end weight and inconsistent loudness between sections — they are not mastered. The raw file often lands well below streaming reference, so after a platform normalises everything to its target, an unmastered AI track can sound noticeably quieter and flatter next to professionally mastered releases. Mastering raises the integrated loudness and evens out the dynamics so it holds up in a playlist.
Is louder always better for streaming?
No. Because the major platforms normalise to a reference level, pushing a master far above the target mostly removes dynamics without making the track louder on playback — it just gets turned down again. The practical goal is to hit roughly the platform reference with a controlled true peak and natural dynamics, not to win a loudness war the normaliser will undo. Results vary by song and genre.
Does loudness mastering change my AI detection score?
Loudness normalisation, EQ and limiting on their own have almost no effect on AI-detection confidence — detectors read spectral fingerprints, not level. Anti-AI Master combines a genre-appropriate loudness master with separate Anti-AI processing that targets those fingerprints to lower the detection confidence score. The two are independent: a great-sounding master does not by itself reduce a detection score, and results vary by song.
One-click loudness-correct mastering
Upload your track for a streaming-ready loudness master, and check your AI detection score free. No account needed.
Open Anti-AI Master Studio →