LUFS targets for YouTube and podcasts are references, not a pass/fail stamp

Loudness is not “make it louder.” Platforms map audio with loudness normalization so a whispery interview and a slammed trailer do not blow a phone speaker apart. The useful measurement is integrated LUFS from an EBU R128-style meter. The Loudness Meter on this site parses FFmpeg’s ebur128 log. It does not apply a limiter. This page is what those numbers mean, and why the target chips are teaching aids.

The three numbers that show up

• Integrated LUFS — the program loudness over the whole file. This is the one people compare to “YouTube −14” or “podcast −16.”
• LRA (loudness range) — how much the program varies. A squashed YouTube ad has a small LRA. A wide film mix has a large one. Neither is automatically “wrong.”
• True peak (when the log includes it) — inter-sample peaks that a sample peak meter can miss. Overs on true peak are how streams distort after encode.

If a true-peak line is missing from the log, the wasm build still measured loudness; it just did not print that field. We do not invent a peak to fill the box.

Why −14, −16, and −23 sit on chips

Creators repeat “YouTube wants −14 LUFS,” “podcasts want −16,” “broadcast wants −23.” Those are useful reference neighborhoods. They are not a certificate this site can issue, and they are not a substitute for the current help docs of YouTube, Spotify, or a network you deliver to.

YouTube has used loudness normalization around the mid −14 LUFS area for many uploads. Podcast hosts often ask for something near −16 integrated with true peak under −1 dBTP. EBU R128 broadcast is −23 LUFS. If your file is −9 integrated, platforms will turn it down. If it is −20, they may turn it up — and also reveal noise you had not heard.

Measuring −14.2 LUFS on this page does not mean YouTube will play it at that level, or that Content ID, ads, or loudness of the source music are “compliant.” It means FFmpeg reported that integrated value for the file you loaded.

Measure first, process elsewhere

A common request is “normalize this to −14.” That is a gain or limiter job. We do not do it here because auto-gain without a true-peak limiter is how people get overs, and a limiter in wasm without review is how people ship pumping artifacts. Use the meter, then adjust in the editor you already have (Resolve, Premiere, Audition, Reaper, a DAW).

Silence removal is a different filter and a different risk: the remover defaults to ends-only so it does not glue speech. Tightening mid-track pauses changes loudness perception even when integrated LUFS barely moves.

When the meter is enough

You exported a podcast episode and want to know if you are in the same neighborhood as last week. You bounced a YouTube voiceover and it “feels quiet” next to other videos — check integrated before you smash a limiter. You received a guest track that might be wildly hotter than your bed. Those are measurement jobs. Certification, broadcast delivery, and mastering are not.