Loudness Targets for Streaming Radio, Explained Plainly

Most of the loudness confusion I see from people prepping tracks for radio comes from treating one number as gospel. There is no single correct loudness target. There are several, and they depend entirely on where the audio is going and whether anything downstream will turn it down for you.

Here is the plain version, the one I wish someone had handed me before I spent a week chasing a number that turned out to be wrong for my station.

The numbers, and what they actually mean

Standardized loudness targets vary by territory and application. These are the ones worth memorizing:

• −24 LUFS: ATSC A/85 (US TV), NPRSS, and PRX radio broadcast
• −23 LUFS: EBU R 128 broadcast, plus Sony’s ASWG standard for console and desktop games
• −19 to −16 LUFS: PRX podcasts
• −14 LUFS: Spotify, Tidal, YouTube, YouTube Shorts, and TikTok
• −27 to −24 LUFS: Netflix, Disney+, and Prime Video

Those come straight from the documented standards, and they are the baseline. Two takeaways matter for anyone running a station or prepping material for one. First, broadcast radio sits around −24 LUFS, much quieter than the streaming services. Second, the streaming platforms sit at −14 LUFS and will pull your track down to that level if you deliver it louder.

The part people miss: normalization is automatic

Loudness normalization works by measuring the average loudness of your audio and applying gain to bring it to the platform’s target. YouTube, for example, prefers −14 LUFS. If your program measures −10 LUFS, YouTube drops it by 4 dB to hit the target. That mechanism exists to stop one song in a playlist from being noticeably quieter than the rest.

So the platform is already doing the level-matching for you. Which raises the obvious question: why master loud at all?

Because loudness normalization does not rescue a bad master. It just rescales it. If you crushed your track to −8 LUFS with heavy limiting to get there, the platform turns it down 6 dB and you are left with a quieter, flatter, more fatiguing version of the same song. You paid for that loudness with dynamics and got nothing back. The normalization gave the advantage to the person who left the transients intact.

Peak normalization is a different beast. On its own it just lifts the highest peak to a set ceiling. But compression and limiting followed by peak normalization is exactly the combination that produces the loudness-war masters we all know. Loudness normalization, the perceptual kind, is the newer and smarter approach because it measures what your ears actually register, not just where the loudest sample lands.

Peaks can still bite you

Here is a trap that catches people delivering to streaming targets. Depending on the dynamic range of the content and the target level, loudness normalization can push peaks past the medium’s limits and cause clipping. Some software offers dynamic range compression to prevent this when it happens, but that alters your signal-to-noise ratio and relative dynamics. In other words, the fix has a cost.

The cleaner answer is true peak limiting on your end. True peak limiting accounts for inter-sample peaks, the ones a sample-accurate meter misses but a playback converter will happily reconstruct into distortion. If your master is headed for a streaming platform that normalizes down, you still want headroom. Leave the ceiling at −1 dBTP and give the normalizer room to work without slamming into the format’s ceiling.

How hard to limit

The old advice was to slap on a limiter and push until it hurt. That approach is dead. Both mastering engineers and the streaming services have pushed back against it, and the practical result is that you want a master that sounds loud but still breathes. A sensible place to begin, for most of what a show throws at you, is the limiter pulling back somewhere between 3 and 4 dB. Watch what it does to transients. Lighter songs need lighter limiting and more transparent attack and release times. Heavier material can survive more aggressive limiting, and a touch of pumping can even work in its favour. But there is a ceiling, and pushing past it kills dynamics and makes the track fatiguing.

Some engineers run two limiters in series rather than one doing all the work. I have done this on dense masters where a single limiter started to smear the kick. Two gentle stages, each doing a couple of dB, kept things cleaner than one stage doing four.

Radio is not Spotify

This is where a lot of internet-radio prep goes wrong. If you are feeding an automated station that applies its own processing (and most do), you are not mastering for the listener’s earbuds. You are mastering for the chain. Delivering a −14 LUFS master into a broadcast chain that expects −24 LUFS gives the station’s processor a hot signal to tame before it has done anything useful. That is not a disaster, but it is not ideal either. Match the delivery target to what the chain expects and let the station’s processing do its job on material that is already in the right neighborhood.

The broadcast number to know is −24 LUFS, which covers ATSC A/85, NPRSS, and PRX radio. If your station publishes a spec, follow it. If it does not, −24 LUFS integrated with a true peak ceiling around −1 dBTP is a defensible default.

Reference, do not guess

Numbers only get you into the right zip code. The rest is reference. Pick songs that became popular within roughly the last two years, lean toward the ones near the top of the charts if you want current sound, and pick several rather than one. Study them before you commit. The worst case is getting an hour into the set before you notice your reference was off and having to begin again.

Compare your master to those references for balance, loudness, and overall feel. If your track measures on target but sounds thin next to the reference, the problem is not the number. It is everything upstream.

The short version

Know your destination. Broadcast radio wants roughly −24 LUFS. Streaming wants −14 LUFS and will normalize you there whether you like it or not. Do not master louder than the target just because you can, because the platform will undo it and leave you with the flattened dynamics. Limit gently, use true peak limiting, leave headroom, and check your work against references that sound like what you are trying to make. Chasing one universal loudness number is a waste of an afternoon. Knowing which number applies to your chain is the whole job.