How normalization works
Streaming platforms measure the integrated loudness of each track (LUFS, ITU-R BS.1770) and adjust the playback volume toward a reference level. If your master is louder than the reference, it is turned down; if it is quieter, some platforms turn it up (only if the peak allows it), others don't. Practical consequence: a master squashed to −6 LUFS on Spotify is turned down by about 8 dB and, at equal volume, can sound smaller and less punchy than a master at −10 LUFS with lively transients. Loudness becomes an aesthetic, genre-based choice, not a race.
The values below are those published or documented at the time of writing; platforms change them: check before an important delivery.
Streaming platforms
| Platform | Reference | Behavior | Recommended true peak | Source |
|---|---|---|---|---|
| Spotify | −14 LUFS (Normal, default); −11 (Loud); −19 (Quiet) | turns down and up; turns up leaving 1 dB of headroom for encoding; album normalization when listening to the album, per-track in playlists and shuffle; in Loud mode applies a limiter (kicks in at −1 dB, attack 5 ms, decay 100 ms) | below −1 dBTP; below −2 dBTP if the master exceeds −14 LUFS | official Spotify for Artists documentation |
| Apple Music (Sound Check) | about −16 LUFS | effectively turns down louder tracks; per track or album | −1 dBTP (Apple Digital Masters) | industry analysis (production-expert, iZotope) |
| YouTube | about −14 LUFS | turns down louder tracks, doesn't turn up quieter ones (observed behavior, not an official specification) | −1 dBTP | industry measurements |
| Amazon Music | about −14 LUFS | normalization on by default | −1 dBTP | industry compilations (iZotope) |
| Tidal | about −14 LUFS | album-based even in playlists | −1 dBTP | industry compilations |
| Deezer | about −15 LUFS | always on, per track | −1 dBTP | industry compilations |
| SoundCloud | conflicting sources (no normalization according to some, about −14 LUFS according to others) | — | −1 dBTP | to be verified |
| Pandora | about −14 (not exactly LUFS) | always on | −1 dBTP | industry compilations |
Social and short-form video (TikTok, Instagram Reels, Shorts)
- No official published reference for TikTok and Instagram; industry measurements indicate adaptive behavior around −14 LUFS. Don't promise a number in a contract.
- The audience often listens through phone speakers: balance matters more than loudness (clear vocal, bass audible through harmonics, no useless sub).
- Practical master: the same master as for streaming (−10/−8 LUFS for pop and hip hop) with a true peak of −1 dBTP works; check it on a phone in mono.
- The first seconds count: the hook must arrive already loud and clear.
Club and DJ
- No normalization: the DJ adjusts the gain; club masters are usually loud (house/techno −9/−6 LUFS integrated, mainstream EDM and hardstyle even −6/−4) but the priority is a mono, stable and powerful low end and clear kick transients.
- True peak: −1/−0.3 dBTP (many DJs use WAV/AIFF files; if you sell on platforms that convert to MP3, keep −1 dBTP).
- Mono sub below 100–150 Hz, no out-of-phase content (club systems often sum the subs to mono).
- Intros and outros with rhythm-only bars (8–32 bars) for mixing, if the genre calls for it.
- Avoid highs that are too harsh: at club volumes 2–5 kHz becomes painful.
CD
- Red Book standard: stereo PCM 16 bit / 44.1 kHz, up to 99 tracks, historical length of 74 minutes (up to about 80 in modern pressing practice).
- No normalization in players: the CD level is the level of the master. A master slightly louder than the streaming one is often used, but it isn't mandatory.
- Peak: −0.3/−0.1 dBFS sample peak (better with a true peak limiter to avoid clipping on cheap players).
- 16-bit dither as the last stage.
- Delivery to the plant as a DDP with PQ, ISRC, CD-Text and EAN/UPC code (see "Stem mastering and delivery formats").
Vinyl
Vinyl has physical constraints: the groove is cut by a stylus, and level, bass, stereo and highs determine space and distortion. Rules of thumb (always ask the cutting engineer for their requirements):
- Duration per side: the longer the side, the lower the cutting level must be. On a 12" at 33⅓ rpm, about 15–20 minutes per side is considered comfortable; beyond 20–22 minutes the level drops noticeably. A 12" at 45 rpm gives more level and quality with shorter sides (about 10–12 minutes).
- Mono bass: phase differences in the bass move the stylus vertically and can make it skip; engineers use an "elliptical" filter that makes the bass below 100–300 Hz mono. It's better to deliver a mix that already has mono bass.
- Sibilance and highs: loud high frequencies (sibilants, cymbals, bright synths above 8–10 kHz) are hard to cut and distort on playback; de-esser and control of the highs.
- Dynamics: you don't need the aggressive limiting used for streaming (in fact, it makes the cut worse). Deliver a less limited version (e.g. −12/−14 LUFS, true peak −1/−3 dBFS depending on the request), at 24 bit.
- Inner-groove distortion: toward the center of the disc the linear speed drops and the highs degrade: put the quieter or less bright tracks at the end of each side.
- Files: 24-bit WAV (44.1, 48 or 96 kHz depending on the mix), one file per side or per track with the side order, gaps and crossfade notes.
Broadcast (TV, radio, video streaming)
- EBU R 128 (Europe): −23 LUFS (±1 LU tolerance where the target isn't practicable, e.g. live), true peak ≤ −1 dBTP; measure with a meter compliant with ITU-R BS.1770 and EBU Tech 3341.
- ATSC A/85 (USA): −24 LKFS.
How to simulate normalization
- Measure the integrated loudness of your master (Emphasis, Youlean).
- Calculate: gain = reference − integrated. Example: master at −8 LUFS on Spotify → −6 dB.
- Lower the output by that amount (Youlean and many meters have platform simulation) and compare with the references normalized the same way.
- If the louder master doesn't sound better than the more dynamic master at equal volume, reduce the limiting.
Practical table by genre (streaming master)
| Genre | Typical integrated | True peak | Note |
|---|---|---|---|
| Trap, rage, drill, phonk | −8/−5 LUFS | −1/−2 dBTP | density and saturation are part of the sound |
| EDM, big room, dubstep | −8/−5 LUFS | −1/−2 dBTP | mono sub, clear kick |
| Pop, K-pop, dance-pop | −10/−7 LUFS | −1/−2 dBTP | vocal up front, strong choruses |
| Boom bap hip hop, R&B | −11/−8 LUFS | −1 dBTP | groove and transients |
| Rock, metal | −10/−7 LUFS | −1/−2 dBTP | avoid harshness |
| Indie, singer-songwriter | −13/−10 LUFS | −1 dBTP | dynamics |
| Jazz, acoustic | −16/−12 LUFS | −1 dBTP | naturalness |
| Orchestral, classical | −20/−14 LUFS | −1 dBTP | wide dynamics (high LRA) |
These are commercial practices, not obligations: choose based on the character you want.
Common mistakes
- Using the streaming master as-is for vinyl.
- True peak at −0.1 dBTP on very loud masters: distortion after AAC/Ogg encoding.
- Chasing −14 LUFS in a genre where the references sit at −7: does the track end up smaller in playlists? Not necessarily: with normalization the difference shrinks; decide by ear, at equal loudness.
- All the tracks of an album at the same integrated loudness: the ballad sounds too loud compared to the single.
Starting settings
| Parameter | Value | Why |
|---|---|---|
| Spotify | −14 LUFS default; TP < −1 dBTP (< −2 if above −14 LUFS) | Official documentation. |
| Apple Music | about −16 LUFS; TP −1 dBTP | Sound Check / Apple Digital Masters. |
| YouTube, Amazon, Tidal | about −14 LUFS; TP −1 dBTP | Industry values. |
| Deezer | about −15 LUFS | Industry values. |
| Club | −9/−6 LUFS (down to −4 in extreme genres), TP −1/−0.3 dBTP | No normalization. |
| CD | 16 bit/44.1 kHz, peak −0.3/−0.1 dBFS, dither | Red Book. |
| Vinyl | Less limited master, 24 bit, mono bass, short sides | Cutting constraints. |
| Broadcast | −23 LUFS (EBU R 128), −24 LKFS (ATSC A/85) | TV/radio standards. |
Frequencies to watch
| Range | What to do |
|---|---|
| Below 100–150 Hz | Club and vinyl: mono bass with no out-of-phase content. |
| 2–5 kHz | Heavily weighted by LUFS (K-weighting): harshness here raises the measured loudness but is fatiguing. |
| Above 8–10 kHz | Vinyl: loud highs and sibilance are hard to cut; streaming: watch out for encoding. |
Example chain
| Step | Tool | Settings | Why |
|---|---|---|---|
| Main master (streaming) | Mastering chain + Emphasis / Pro-L 2 | Genre loudness, −1 dBTP | Reference version. |
| Normalization simulation | Youlean Loudness Meter / Emphasis + gain | Lower to the platform's reference | Realistic comparison. |
| Club version (if different) | Same chain, controlled low end | TP −1/−0.3 dBTP | Maximum impact without normalization. |
| CD version | 16-bit dither + DDP | Peak −0.3 dBFS | Pressing. |
| Vinyl version | Chain with reduced or no limiter | −12/−14 LUFS, 24 bit | Clean cut. |
How to check the result
- For each version: note the integrated loudness, maximum short-term, true peak and LRA.
- Streaming: true peak ≤ −1 dBTP (≤ −2 dBTP if above −14 LUFS).
- Spotify/Apple simulation: the normalized master holds up against the references normalized the same way.
- Vinyl: correlation below 150 Hz > +0.9, no excessive sibilance above 8 kHz, duration per side noted.
- CD: sample peak ≤ −0.1 dBFS, 16 bit with dither, no clipping.