LUFS targets for YouTube, Spotify and podcasts, explained

Why louder doesn't win any more

For years, the way to stand out on the radio or in a playlist was to be louder than everything around you. Mixes were squeezed harder and harder until the quiet parts disappeared and the loud parts distorted. Then streaming services started measuring how loud each track and episode sounds and adjusting playback so everything lands at a similar level.

That changes the game completely. A master pushed far above a platform's reference gets turned down, and all it keeps is the damage: flattened dynamics and a tiring, crushed sound. A master that's far too quiet has the opposite problem, it can sound weak next to the content before and after it. Hitting a sensible loudness target is now about sounding consistent and clean, not about winning a volume contest.

LUFS and true peak in plain words

LUFS (loudness units relative to full scale) is the measurement platforms use for perceived loudness, based on the ITU-R BS.1770 standard. Rather than looking at the highest point of the waveform, it filters the audio to mimic how our ears weigh different frequencies and averages the result over time. The integrated value covers the whole file and ignores near-silent stretches, so pauses in a podcast don't drag the number down. A change of one LU equals a change of one decibel, so a file at −18 LUFS needs 4 dB of gain to reach −14.

True peak (measured in dBTP) is the highest level the audio reaches once it's turned back into a continuous waveform, including peaks that fall between digital samples, which a simple sample-peak meter misses. It matters because lossy encoding to MP3, AAC or Ogg Vorbis can push those in-between peaks higher. A file that peaks right at 0 dB before encoding may clip after it, which is why platforms ask you to leave a little headroom.

What each platform documents

Spotify

Spotify's artist documentation is the most detailed. At the time of writing it says:

• Playback is normalized to −14 LUFS, measured to the ITU 1770 standard. Your file isn't altered; the gain is applied as it plays.
• Louder masters are simply turned down, with no added distortion. Quieter masters are turned up only as far as their peaks allow: Spotify keeps about 1 dB of headroom for lossy formats, so a −20 LUFS track that peaks at −5 dBTP is only raised to −16 LUFS.
• Play a whole album and it's normalized as one unit, so quiet songs stay quiet relative to loud ones. On shuffle or in a playlist, each track is adjusted on its own.
• Premium listeners can choose Loud (−11 LUFS, with a limiter), Normal (−14) or Quiet (−19).
• The web player and third-party devices such as TVs and speakers don't apply normalization, so your master's own level still reaches some listeners untouched.

Spotify's mastering advice follows from that: aim for −14 LUFS integrated with true peaks below −1 dBTP, and if you master louder than −14, keep peaks below −2 dBTP, because loud tracks are more prone to distortion when they're encoded for streaming.

Apple Podcasts

Apple's podcast audio requirements recommend an integrated loudness of around −16 LKFS, within ±1 dB, with true peaks kept at −1 dBFS or lower (LKFS is the same unit as LUFS under another name). Apple also asks for this to be done before encoding, since the encoder won't fix loudness for you and may clip a file whose peaks are too hot.

YouTube

YouTube doesn't publish a loudness target in its help center at the time of writing. What it does document is Stable volume: a viewer-side feature, on by default, that continuously evens out the gap between quiet and loud passages. It isn't available for every video, it's switched off for YouTube Music and official music videos, and viewers can turn it off. In other words, you can't rely on YouTube to rescue a badly levelled mix, and you shouldn't try to out-shout it either.

So which number should you use?

• Music for Spotify and other streaming services: an integrated −14 LUFS, peaking no higher than −1 dBTP. That matches Spotify's published advice exactly.
• Podcasts: −16 LUFS (Apple's −16 ±1) and a −1 dBTP peak ceiling. Spotify doesn't publish a separate podcast loudness spec in its help pages at the time of writing, so Apple's documented figure is the safer common ground for a single master.
• YouTube videos: with no official number to hit, keep the integrated level somewhere around −14 to −16 LUFS with true peaks under −1 dBTP, and, more importantly, keep dialogue consistent from start to finish and music beds clearly below the voice.
• One file for everywhere: pick by content, −16 for speech-led content, −14 for music-led content, and always respect the −1 dBTP ceiling.

Treat these as targets to land near, not exact figures. Within about a decibel is close enough for any platform that normalizes, and a natural, uncrushed sound matters far more than the last tenth of a LU.

Measure first, then fix

1. Measure the finished mix. Open the file in the loudness meter. It reads integrated loudness, true peak and loudness range, plus the loudest short-term and momentary moments, and draws a graph of loudness over time.
2. Fix uneven sections before normalizing. A quiet guest or a loud laugh won't be balanced by a single gain change. Lower or raise those stretches with Volume in the audio editor, and cut long pauses while you're there.
3. Pick a target and a ceiling. Choose a preset or type a custom LUFS value, and set the true-peak ceiling (−1 dBTP covers the platforms above). Before anything is applied, you'll see the exact gain change and whether the limiter will have any work to do.
4. Normalize and check the result. The level is changed with one gain move, a limiter catches only the peaks that would cross your ceiling, and the saved file is measured again so the after-numbers describe what you actually download.

Only need the gain change? Normalize Audio opens the same tool ready to go. For a video, the soundtrack is measured straight from the file and saved as audio, so bring it back into your editor afterwards, or pull the sound out first with Extract Audio.

Mistakes that cost quality

• Peak-normalizing instead of loudness-normalizing. Raising a file until its highest peak hits 0 dB says nothing about how loud it sounds, and leaves no headroom for encoding.
• Forcing a quiet, dynamic recording up to −14. If reaching the target takes many decibels of limiting, the audio flattens out. A slightly lower target, or evening out the loudest moments first, sounds better.
• Mastering far louder than the target. It gets turned down anyway; the only lasting effect is the lost punch.
• Measuring a short excerpt. Integrated loudness describes the whole programme. Measure the finished file, not a 30-second section.
• Forgetting that encoding moves peaks. Exporting to MP3 or AAC after normalizing can nudge true peaks up slightly. Re-measure the encoded file, or save WAV for strict deliveries.