How loudness is measured
The reference standard is ITU-R BS.1770. The measurement applies K-weighting (a filter that attenuates low frequencies and slightly raises the highs above 2 kHz, imitating the sensitivity of the ear), calculates the average power and expresses it in LUFS (Loudness Units relative to Full Scale; LKFS is the same value under another name). 1 LU = 1 dB.
| Measurement | Window | Use |
|---|
| Momentary (M) | 400 ms | Instantaneous response |
| Short-term (S) | 3 s | Loudness of sections (chorus, drop) |
| Integrated (I) | whole track, with gating | Normalization value |
| LRA (Loudness Range) | distribution of the short-term values | Variation between loud and quiet parts |
| True peak (dBTP) | peak reconstructed with oversampling | Inter-sample peaks after D/A conversion or encoding |
| PLR (Peak to Loudness Ratio) | true peak − integrated | Indicator of transient dynamics |
Gating of the integrated measurement: anything below −70 LUFS (absolute threshold) and anything more than 10 LU below the provisional integrated value (relative threshold) is ignored. Silence and very quiet parts do not lower the result.
Broadcast and video standards
| Standard / platform | Level | Maximum true peak | Notes |
|---|
| EBU R128 (European TV and radio) | −23 LUFS | −1 dBTP | Tolerance ±0.5 LU; ±1 LU for live programs |
| ATSC A/85 (US TV) | −24 LKFS | −2 dBTP | Tolerance ±2 dB |
| Netflix | −27 LKFS measured on speech (dialog gating) | −2 dBTP | Tolerance ±2 LU |
| Podcasts (common practice of Apple Podcasts and Spotify) | −16 LUFS (stereo) | −1 dBTP | Speech and background music |
In broadcast, loudness must be met, it is not a ceiling: a program at −20 LUFS on an R128 channel is corrected or rejected.
Music streaming normalization
| Service | Playback level |
|---|
| Spotify | −14 LUFS (Loud mode −11, Quiet −19) |
| YouTube | −14 LUFS (turns down louder tracks, does not turn up quieter ones) |
| Apple Music (Sound Check) | −16 LUFS |
| Amazon Music, Tidal | −14 LUFS |
| Qobuz | −18 LUFS |
Normalization only changes the playback volume: a master at −8 LUFS on Spotify is turned down by 6 dB. The louder master does not sound louder, it just sounds less dynamic. Platforms recommend a true peak of −1 dBTP (−2 dBTP for masters louder than −14 LUFS) because lossy encoding (AAC, Ogg Vorbis, Opus) adds peaks.
Typical LRA
| Material | LRA |
|---|
| EDM, trap, modern pop | 3–6 LU |
| Rock, hip hop | 4–8 LU |
| Acoustic pop, jazz | 6–12 LU |
| Film score, classical | 10–20 LU |
| TV programs (EBU recommends adapting to the context) | 5–15 LU |
Youlean Loudness Meter (free), iZotope Insight and Ozone, Waves WLM, FabFilter Pro-L (integrated measurement), NUGEN VisLM, FL Studio's Wave Candy (peak and spectrum). The app measures integrated, maximum short-term, sample peak (not true peak), crest factor and correlation.
Procedure
- Measure the whole track from start to finish (integrated) and note the maximum short-term and LRA.
- Check the true peak with oversampling (4× or more).
- For broadcast: bring the integrated loudness to the target with a constant gain, then check the true peak; if it is too high, use a true peak limiter with a −2/−1 dBTP ceiling.
- For streaming: choose the loudness the genre needs, then check how much the platform will turn the track down.
Common mistakes
- Confusing sample peak and true peak: a file with a sample peak of −0.1 dBFS can have a true peak of +1 dBTP.
- Measuring only the chorus and calling it “integrated.”
- Delivering a music mix at −10 LUFS to a TV channel: it must be brought back to −23 or −24 according to the channel's rules.
- Chasing −14 LUFS for streaming at the expense of the genre's impact: the normalization target is not a mandatory mastering target.
Starting settings
| Parameter | Value | Why |
|---|
| EBU R128 | −23 LUFS ±0.5 LU, −1 dBTP | European TV and radio standard. |
| ATSC A/85 | −24 LKFS ±2, −2 dBTP | US TV standard. |
| Netflix | −27 LKFS dialog-gated, −2 dBTP | Streaming video content. |
| Spotify, YouTube, Tidal, Amazon | −14 LUFS | Normalization level. |
| Apple Music | −16 LUFS | Sound Check. |
| True peak for streaming | −1 dBTP (−2 for loud masters) | Margin for lossy encoding. |
| True peak meter oversampling | 4× or more | Correct measurement of inter-sample peaks. |
Frequencies to watch
| Range | What to do |
|---|
| Below 100 Hz | K-weighting gives less weight to the bass: a lot of sub raises the LUFS only slightly but uses up headroom. |
| 2–6 kHz | K-weighting gives more weight to this region: the high mids make the measured loudness rise. |
Example chain
| Step | Tool | Settings | Why |
|---|
| Mastering | Full chain | Genre loudness | Final sound. |
| Measurement | Youlean / Insight / WLM | Integrated, short-term, LRA, true peak | Delivery data. |
| Broadcast normalization | Constant gain | Standard's target | Compliance with the specification. |
| True peak limiter | Pro-L / Maximus / broadcast limiter | Ceiling −2/−1 dBTP | Compliant peaks. |
| Final check | New measurement of the exported file | All values | Compliance. |
How to check the result
- Integrated loudness measured on the whole file, not on a section.
- Maximum short-term no more than 3–5 LU above the integrated for pop and electronic music; for broadcast, no constraint but an LRA suited to the content.
- The sample peak shown by the app is a minimum: leave 0.5–1 dB of margin relative to the required true peak, or check with a true peak meter.
- For broadcast: integrated within tolerance (R128 ±0.5 LU, ATSC ±2 dB).