LUFS targets for streaming and podcasts

The short answer

LUFS (Loudness Units relative to Full Scale) is how a platform measures the perceived loudness of a whole program, averaged over time. Each place your audio can land has a rough integrated target: music streaming aims for about -14 LUFS, podcasts and spoken word for about -16 LUFS, and broadcast for -23 or -24. The scale is negative because 0 LUFS is the full-scale ceiling and real program material always sits well below it.

The single most useful thing to understand is that you do not need to be loud. On-demand platforms measure your finished program and turn loud masters back down to their target, so pushing past the number buys you nothing but squashed dynamics. The job there is simple: match the target level, protect the peaks with a ceiling, and stop. Live platforms like Twitch and YouTube are the exception, they normalize nothing, and we cover what that changes below.

The targets, by where your audio is going

There is no single correct number, only the number for the destination, and the first split is whether that destination is on demand or live. An on-demand platform measures your finished file and normalizes it to a target; a live platform sends your audio out as it arrives and normalizes nothing. These are the common integrated LUFS targets, each paired with a true-peak ceiling to deliver underneath.

• Music streaming: about -14 LUFS. On-demand services such as Spotify, Tidal, Amazon Music, and YouTube normalize toward roughly -14 LUFS integrated. A few sit a little different (Apple Music references about -16), but master to -14 and you are inside the window nearly everywhere that matters.
• Live streaming, Twitch and YouTube Live: aim for about -14 LUFS. Twitch does not publish a loudness target and applies no normalization of its own, which is exactly why one channel blasts you and the next is a whisper. YouTube turns loud uploads down on playback, which lands them near -14, but is looser when you go live. With nothing correcting you in the moment, treat -14 LUFS integrated under a -1 dBTP ceiling as a sensible aim you hold yourself, watched live on a loudness meter in your streaming app.
• Podcasts and spoken word: about -16 LUFS. Apple Podcasts references -16 LUFS, and it is the safe default for talk. Some platforms use -14 instead, so check your host. One nuance: a mono file and a stereo file measured at the same LUFS are equally loud, but many podcast hosts want a specific channel count, so pick the format your platform expects and measure that file.
• EBU R128 broadcast: -23 LUFS. The European broadcast standard, with a tolerance of about plus or minus 0.5 to 1 LU. Much quieter than streaming because broadcast leaves real headroom for dynamics.
• ATSC A/85: -24 LKFS. The US broadcast standard. LKFS is the same unit as LUFS under a different name, so -24 LKFS is simply one unit quieter than EBU R128.

Pair every target with a true-peak ceiling

A LUFS target controls your average loudness and says nothing about your single tallest peak, so it is only half the delivery spec. The other half is a true-peak ceiling. When digital samples are reconstructed into an analog waveform on playback, the smooth curve between two samples can rise slightly higher than either sample, an inter-sample peak that a plain sample meter never sees.

Deliver with a true-peak ceiling around -1 dBTP, and drop to about -2 dBTP for low-bitrate lossy formats where lossy encoding adds a little extra overshoot. That headroom keeps a program that measures legal actually staying legal after the platform converts and re-encodes it. Peak and loudness are separate axes: two programs can share the same -0.1 dBFS peak yet differ by many LU, which is exactly why you set them independently.

How to actually hit a target

Hitting an integrated target is a measure-and-move loop on the finished program, not something you eyeball while mixing. The order matters.

1. Finish the mix first, with clean gain staging through the chain so nothing is fighting for headroom before you start measuring.
2. Measure the integrated LUFS of the whole finished program on a loudness meter. Integrated uses gates (silence does not drag the number down), so it reflects the loudness a listener actually hears across the piece.
3. Compare to your target and move the overall level. Because 1 LU equals 1 dB, the correction is just subtraction.
4. Put a limiter last to hold the peak ceiling, set around -1 dB. Use it as a safety net for stray peaks, not as a loudness tool. If you are limiting more than a dB or two just to reach a LUFS target, you are too hot: turn the whole thing down instead.
5. Re-measure the finished file and confirm both numbers, the integrated LUFS and the true peak, land where you want them.

Why you cannot cheat the target

It is tempting to master a few LU hotter than the target, hoping to sound louder than the next program. You cannot. Loudness normalization means the playback system measures your program's integrated LUFS and adjusts its own playback gain to the common target, turning loud masters down (and often quiet ones up). A hyper-compressed master does not come out louder; it comes out at the same target as everyone else, only with its dynamics already crushed for no reward.

That is the whole reason the loudness war ended. Once the destination normalizes everything to one level, being louder at the source is a self-inflicted wound. Match the target, keep your dynamics, and let the limiter guard the ceiling. Louder is not a strategy anymore; it is just a worse-sounding way to arrive at the identical loudness.

Live streaming is a different game

Everything above assumes a finished file the platform can measure and normalize. A live stream has no finished file. On Twitch, and largely on a live YouTube broadcast, whatever loudness leaves your machine is what your viewers hear, with no second pass to even it out. That changes the job in two ways.

First, consistency is on you. Since nothing normalizes the stream, a viewer who set their volume during your intro should not get blasted when the game explosion hits, or strain to hear you under the music. Aim for about -14 LUFS integrated across the stream, keep a limiter on the output as a safety net for spikes, and watch a loudness meter as you go rather than after.

Second, and more important for a game streamer, the number that decides whether you sound good is not the master LUFS at all: it is the balance between your sources. Your voice has to sit clearly above the game, the music, and any voice chat, and keeping those on separate channels is what makes that balance adjustable in the moment rather than baked into one muddy mix. Get the relative levels right and a sane overall loudness tends to follow.

LUFS in patchd

Here is the honest scope. patchd is a live mixer and monitoring tool, not a mastering loudness meter, so it does not measure LUFS. Its channel meters show peak level in dBFS: a live tracked dB readout per channel from -60 up to 0, with silence shown as negative infinity, plus a VU meter fused into the fader. That is instantaneous peak and level, which is a different measurement from program loudness. LUFS is a whole-program number that the streaming or podcast platform computes downstream on your finished file.

What patchd gives you toward these targets is the clean signal to hand off. Good gain staging on every channel means your audio reaches the output at a sane level with real headroom, and the free limiter node holds a clean peak ceiling on the way out at its default -1 dB, with zero added latency. One honest caveat: because that limiter has no lookahead, it catches sample peaks, not true peaks (true-peak limiting needs the lookahead and oversampling a live monitoring path deliberately avoids). So treat it as peak safety for your live signal, and let the recording or streaming platform do the final LUFS normalization and true-peak check on the finished program. The metering and the limiter are free, and the real-time engine is identical on both tiers.

For a live stream this is where a mixer earns its place, because the live path is the whole job. Each source, your mic, the game, music, and voice chat, sits on its own channel with its own fader and peak meter, so you set the balance and the overall level in real time and route the result into OBS or straight to your streaming app. There is no finished file and no platform normalization to lean on, so the mix you send is the mix that airs, and a per-channel mixer is exactly the tool for holding it steady.

The takeaway

Aim for the number that fits the destination: about -14 LUFS for music streaming, about -16 LUFS for podcasts and spoken word, and -23 or -24 for broadcast, each delivered under a true-peak ceiling near -1 dBTP (-2 for low-bitrate lossy). Hit it by measuring the integrated LUFS of the finished program and moving the overall level (1 LU = 1 dB), using a limiter only to hold the peak ceiling. Do not exceed the target hoping to be louder, because normalization cancels it and keeps only the damage. And when you go live on Twitch or YouTube, remember nothing normalizes you at all: hold about -14 yourself, and spend your effort on the balance between your voice and everything else.

patchd is a pre-launch Windows audio mixer with clean per-channel metering and a free limiter, built to hand your platform a tidy, headroom-safe signal so its normalization has an easy job. patchd is in development now. Join the waitlist and we will tell you the moment it is ready to install.