Peak: the loudest single instant
Peak is the simplest measurement and the least useful on its own, since it only reports the single highest point the waveform reaches. Two files can share an identical peak and sound completely different in overall loudness, since peak says nothing about everything else in between.
RMS: the average level
RMS (root mean square) averages the signal's level over time, which represents loudness far better than peak alone. It's still a raw measurement of the waveform rather than something tuned to human hearing. ACX specifies RMS directly for audiobook submissions, requiring each file to sit between -23 dB and -18 dB RMS.
RMS is especially useful for audiobooks because dynamic range is crucial to a good listener experience. In a three-minute song the general loudness stays fairly similar throughout, whereas in a book there might be whole five-minute sections that sit at different levels for dramatic effect. An RMS window makes sure the overall level stays consistent without flattening those intentional shifts.
LUFS: loudness tuned to how you actually hear
LUFS (Loudness Units relative to Full Scale) weights the signal by frequency to match how sensitive human hearing actually is at each one, then averages that over time. It's the standard modern streaming and podcast platforms use, because they're normalising perceived loudness across a huge catalogue of very different content, from music to speech to sound effects, where a raw RMS number wouldn't compare fairly. It's particularly helpful for music platforms, where -14 LUFS has become the industry standard.
There's been an interesting phenomenon with LUFS over the last twenty years or so: songs have been getting louder and louder, in what's known as the loudness war. As streaming services and artists fight for our attention span, a louder song pushes its harmonies further into your bloodstream, so artists and engineers have been edging the LUFS up year by year. Songs from the 1970s were much quieter in that sense than modern music, which uses far more limiting and compression to get there.
Why different platforms specify different ones
Audiobook platforms check individual files against fixed technical limits, so RMS and peak make sense; they're simple, objective and easy to enforce. Podcast and streaming platforms are normalising playback across millions of unrelated files in real time, so LUFS, built specifically to estimate perceived loudness, is the better tool for that job.
Quick reference
- Audiobooks (ACX): -23 dB to -18 dB RMS, with peak values no higher than -3 dB.
- Podcasts: often mastered using LUFS-based targets rather than RMS; the exact target depends on the workflow and delivery platform.
- Streaming/video platforms: commonly apply their own playback loudness normalisation, so their loudness behaviour should not be treated as an ACX mastering target.
Preparing an audiobook for ACX? Read our full guide to ACX loudness requirements, including how RMS, peaks and noise floor affect each other during mastering.
If you're mastering narration for audiobook delivery, see how Optivox handles audiobook mastering and ACX-style targets.
Optivox picks the right target for your recording
Which number you need depends on where the file is going, and Optivox handles that for you. Podcasts and general recordings get LUFS-based mastering, while Audiobook mode masters specifically to ACX's RMS window and peak limit.