Blog · Color and sound
Loudness targets by platform
· 5 min read
Short answer
As of November 2026, Spotify publishes -14 LUFS, Apple Podcasts -16 LKFS, Netflix -27 LKFS dialogue-gated, EBU R128 -23 LUFS and ATSC A/85 -24 LKFS. YouTube, TikTok and Instagram publish no figure on the pages we read. Consumer services change playback gain when you miss; delivery specs can fail QC.
On this page
What each one publishes, as of November 2026
Every figure below was read on the publisher’s own page on November 4, 2026. “None found” means the pages listed in the sources give no number, not that the service does nothing.
| Service or standard | Target | Measured | True peak | When you miss |
|---|---|---|---|---|
| YouTube | None found | Not stated | Not stated | May adjust levels and mix automatically. Stable volume is on by default for viewers |
| Spotify, music | -14 LUFS. Listener settings: Loud -11, Quiet -19 | ITU 1770, integrated | Under -1 dB; under -2 dB if the master is above -14 LUFS | Louder is turned down. Quieter is raised, but only as far as 1 dB of headroom allows |
| Apple Podcasts | -16 LKFS, ±1 dB | BS.1770-5 | -1 dBFS | A recommendation; no penalty stated. Warns that encoding may clip the signal |
| Apple Music | None on its Sound Check page | Not stated | Not stated | Sound Check raises or lowers each track, if the listener has it on |
| Netflix, branded | -27 LKFS, ±2 LU | BS.1770-1, dialogue-gated, full program | -2 dB | QC partners flag it as an issue past the wider thresholds below |
| EBU R128, European broadcast | -23 LUFS. ±1 LU only where the target isn’t practical, such as live | Whole program, gated | -1 dBTP in production | A recommendation; each broadcaster enforces it. Allows ±0.2 LU for measurement error in QC |
| ATSC A/85, US TV | -24 LKFS without metadata or a prior agreement; about ±2 dB of measurement tolerance | Dialogue for long form, full mix for short form | -2 dBTP | A recommended practice. The FCC rule built on it covers TV commercials only |
| Prime Video ads | -24 LKFS, ±2 dB | Not stated | -2 dB | Not stated on the spec page |
| TikTok, Instagram Reels | None found on their ad specs | Not stated | Not stated | Not stated |
Four more that don’t fit a row. Apple’s Podcasts page says content with Sound Check metadata plays at -16 dB when the setting is on, while its Apple Music pages give no number. Apple’s asset guide (v5.3.19) gives a range for film in Dolby Atmos, dialogue between -31 and -10 LKFS, rather than a target. Prime Video’s content guide for licensors, last edited in 2018, has no loudness figure at all. Spotify’s video specs don’t mention loudness either, so the -14 above is from its music pages.
A/85 was revised in July 2026, and the new text adds a range for streaming services: a single target between -23 and -27 LKFS. EBU R128 s2 says to stream programs unchanged at -23 LUFS, with -20 to -16 LUFS as an interim level where no metadata manages the device’s gain.
Does YouTube have a loudness target?
Not one it publishes. YouTube’s upload settings name audio codecs and no level. Its volume page describes Stable volume, which keeps adjusting levels to reduce the gap between quiet and loud. That’s off for YouTube Music and official music videos, and not every video has it.
You can opt a channel out of audio enhancements for future uploads, in YouTube Studio under Settings, Channel, Advanced settings. YouTube warns that viewers may then lose those features. The -14 LUFS you’ll see attributed to YouTube isn’t on any of these pages. Treat it as someone’s measurement.
What actually happens when you miss
There are three different outcomes, and they call for different caution.
The service turns you down, or up. Spotify’s numbers show how this plays out. A master at -10 LUFS comes down 4 dB at the default setting and 9 dB at Quiet, so the loud master plays no louder and has less range. Going the other way, Spotify’s own example is a track at -20 LUFS with a -5 dB true peak: it comes up only to -16 LUFS, because it stops 1 dB short of the ceiling.
A person or a QC tool flags it. Netflix publishes both numbers, which is rare. The spec says -27 ±2 LU and -2 dB true peak. Its QC guidance flags a mix only if it fails -27 ±3 dialogue-gated and then also fails -24 ±3 on full program, and it holds print-master peaks to -1 dB. Those wider limits exist to absorb differences between meters, and the page says the spec itself hasn’t moved. Mix to the spec.
Nothing automatic happens. R128 and A/85 are recommendations until a broadcaster writes them into a contract. The CALM Act rule is narrower than people assume: the FCC says it requires commercials to match the average volume of the programs around them, sets no absolute cap, doesn’t cover the internet, and relies on complaints to spot patterns.
Why the same mix gives two readings
Because the meters listen to different things. R128 measures everything in the program. A/85 anchors long form on dialogue, and short form on the full mix, and Netflix gates on dialogue. A film with quiet dialogue and a loud score reads lower on a dialogue gate than on the full mix, so a full-mix meter can’t confirm a dialogue-gated spec.
The BS.1770 version matters as well. A/85 and Netflix both name BS.1770-1 for dialogue-gated long form, because later versions added a relative gate. If a spec names a version, set your meter to it.
Which number should you use?
Use the delivery spec if there is one, and read the current one, since these change. A/85 changed this year. Without a spec, this is our judgment, not a standard:
- Online, one master, no spec: -16 to -14 LUFS, true peak at -1 dB or lower. That sits on the two published consumer figures and the top of R128 s2’s interim range.
- Broadcast: the broadcaster’s sheet, then -23 LUFS in Europe and -24 LKFS in the US.
- Streaming series or film: the service’s spec, with a dialogue-gated meter if it asks for one.
Render separate masters rather than one compromise. Arithmetic is why: a mix at -24 LKFS with peaks at -2 dBTP has 1 dB of room before it reaches -1 dB. Bringing it to -16 takes 8 dB, so about 7 dB of peaks have to go through a limiter. Listen to that version, on a phone as well.
To measure a file with FFmpeg’s ebur128 filter (we ran this on FFmpeg 8.1.1):
ffmpeg -i mix.wav -af ebur128=peak=true -f null -
The summary at the end gives integrated loudness, loudness range and true peak. It reads the whole mix, with no dialogue gate. It’s one of several checks worth running before you deliver.
Setting a target on delivery in nolanlabs
nolanlabs, the editor we make, takes a target loudness and a peak on delivery. It measures the mix by EBU R128 and applies plain gain when that stays under the peak, so the mix keeps its shape. Otherwise it uses loudnorm’s second pass. If the target can’t be reached under the peak, the peak wins, and the check after the render shows how loud it actually got, with true peak and loudness range. That measurement covers the whole mix, so use a dialogue-gated meter for a Netflix spec. The mix order is on the sound page.