LUFS, true peak, and loudness range answer different questions. Use them together to describe a render, then compare the result with your delivery brief and your reference track.
Measurement and targets
EBU R 128 recommends an average programme loudness of -23 LUFS for its broadcast context, alongside loudness range and maximum true peak descriptors. That recommendation does not make -23 LUFS the required level for every music release.
Phantom’s loudness tool reports integrated, short-term, and momentary loudness, loudness range, and true peak. Integrated loudness summarizes the program; short-term and momentary values describe shorter windows. A peak measurement helps check headroom, while loudness range describes variation over time.
Sample peak and true peak are also distinct. Sample peak describes stored values. True peak estimates the reconstructed waveform between them. A float WAV can preserve samples above 0 dBFS, so check levels before converting to an integer delivery format.
Playback normalization is a separate step
Spotify’s normalization guide describes -14 LUFS for its Normal playback setting. It also documents other listener settings and devices without normalization. The adjustment happens during playback; it is not a universal upload rejection threshold.
Spotify recommends a maximum true peak below -1 dBTP, or below -2 dBTP for masters louder than -14 LUFS, to leave room for encoding. Read the current platform or distributor instructions for your delivery. Avoid treating one service’s advice as a rule for every destination.
Check a render with Phantom
Ask your assistant:
Measure the loudness of mix-v3.wav. Report integrated LUFS, true peak, and loudness range. Compare true peak with the -1 dBTP ceiling in my delivery brief.
The ceiling in that prompt is a chosen requirement. analyze_loudness returns measurements; the assistant supplies the comparison and explanation.
For example, a render at -14.2 LUFS and -0.8 dBTP exceeds that chosen ceiling by 0.2 dB. Reducing gain can address the level. It cannot restore distortion already printed into the source. Re-measure the new file, and audition both versions at matched playback levels before keeping a processing change.
Keep the comparison useful
- Save an unprocessed render and its measurements.
- Make one deliberate change, then render a new file.
- Compare loudness, peaks, and dynamics with the first render.
- Listen at matched levels for distortion, lost transients, and changes in balance.
Use reference comparison for a broader view of spectrum, dynamics, and stereo differences. A closer numerical match is helpful when it supports the sound you want.