Hear your video the way each device plays it.

Drop the file. Listen through seven devices, from a phone speaker to a car, and read its integrated loudness, true peak and loudness range against YouTube's target. Measured in your browser.

The sample on seven devices · dB relative to over-ear headphones
  1. Phone speaker−11.1 dB54% of the mix is below what a phone speaker can play.
  2. Laptop speakers−4.5 dB
  3. TV built-in−3.6 dB
  4. Soundbar−1.2 dB
  5. Earbuds+0.5 dB
  6. Over-ear headphones0 dB
  7. Car−0.5 dB

Before any of these, YouTube turns the whole file down 5.2 dB: it is -8.8 LUFS against a -14 target.

Check your video

Have a subscription key?

sample-video.mp40:22 · stereo · 48 kHzwhole file analysed

Integrated loudness
-8.8 LUFS

YouTube plays this 5.2 dB quieter than you hear it now (target -14).

True peak
-0.5 dBTP

Above the −1 dBTP ceiling: lossy encoding can clip the loudest moments.

Loudness range
12.4 LU

Quietest passage -25.1, loudest -6.6 LUFS — after YouTube's turn-down the quietest lands near -30.3.

Short-term loudness, 3-second window, and the target line0-10-20-30-40-14 LUFS targetmusic entersvoice0:00 · voice over bed -1.6 dB0:01 · voice over bed -5.6 dB0:02 · voice over bed -5.2 dB0:03 · voice over bed -4.7 dB0:04 · voice over bed -2.1 dB0:05 · voice over bed -5.0 dB0:06 · voice over bed -4.3 dB0:07 · voice over bed -4.1 dB0:08 · voice over bed -4.2 dB0:09 · voice over bed -4.2 dB0:10 · voice over bed -4.1 dB0:11 · voice over bed -4.2 dB0:12 · voice over bed -4.2 dB0:13 · voice over bed -4.0 dB0:14 · voice over bed -4.1 dB0:15 · voice over bed -4.1 dB0:16 · voice over bed -4.1 dB0:17 · voice over bed -4.2 dB0:18 · voice over bed -4.2 dB0:19 · voice over bed -4.1 dB0:20 · voice over bed -4.1 dB0:21 · voice over bed -4.1 dB0:000:050:110:160:22
Short-term loudness, 3-second window, and the target line0-10-20-30-40-14 LUFS targetmusic entersvoice0:00 · voice over bed -1.6 dB0:01 · voice over bed -5.6 dB0:02 · voice over bed -5.2 dB0:03 · voice over bed -4.7 dB0:04 · voice over bed -2.1 dB0:05 · voice over bed -5.0 dB0:06 · voice over bed -4.3 dB0:07 · voice over bed -4.1 dB0:08 · voice over bed -4.2 dB0:09 · voice over bed -4.2 dB0:10 · voice over bed -4.1 dB0:11 · voice over bed -4.2 dB0:12 · voice over bed -4.2 dB0:13 · voice over bed -4.0 dB0:14 · voice over bed -4.1 dB0:15 · voice over bed -4.1 dB0:16 · voice over bed -4.1 dB0:17 · voice over bed -4.2 dB0:18 · voice over bed -4.2 dB0:19 · voice over bed -4.1 dB0:20 · voice over bed -4.1 dB0:21 · voice over bed -4.1 dB0:000:110:22
Short-term loudness (3 s) over time. The strip below is the voice-over-bed estimate: darker where the voice sits further under the music.

Phone speaker

Mono, nothing below ~300 Hz, a bump around 3 kHz. · modelled, not measured

Plays at
-19.9 LUFS-11.1 dB vs headphones · integrated, after this device's response
Below its 300 Hz floor
54 %of the original mix's energy is below what this device can play
Voice over bed · estimate
5.2 dBdips to 3.4 · energy in 3003400 Hz against the rest, median over the file; wrong for sung vocals or very close-miked voices
All seven devices, compared
On each device — modelled, not measured
DevicePlays atBelow its floorVoice over bedListen
Phone speakerMono, nothing below ~300 Hz, a bump around 3 kHz.-19.9 LUFS-11.1 dB vs headphones54 %5.2 dB
Laptop speakersFloor around 180 Hz, a little presence, narrow stereo.-13.3 LUFS-4.5 dB vs headphones39 %1.7 dB
TV built-inFloor ~120 Hz, thin low-mids, brighter top.-12.4 LUFS-3.6 dB vs headphones26 %-0.8 dB
SoundbarReaches ~60 Hz, lifted lows, slightly recessed mids.-10.0 LUFS-1.2 dB vs headphones7 %-5.2 dB
EarbudsNear-flat, a low-shelf lift, the ear-canal bump at 3 kHz.-8.3 LUFS+0.5 dB vs headphones4 %-4.6 dB
Over-ear headphonesFlat. Probably what the mix was made on.-8.8 LUFS1 %-4.0 dB
CarBig low shelf, mids pulled back, top rolled off.-9.3 LUFS-0.5 dB vs headphones4 %-6.3 dB

Plays at is integrated loudness after each device's response. Below its floor is the share of the original mix's energy under that device's low-frequency floor. Voice over bed is an estimate (speech band against the rest, median over the file).

Where the energy is, by band
Sub7%2060
Low28%60150
Low-mid19%150300
Voice45%3003400
Presence0%34008000
Air0%800016000
  1. 1Lower the master by about 5.2 dB.

    Integrated loudness is -8.8 LUFS; YouTube normalises to -14, so it will play 5.2 dB quieter than you hear it now — including the quiet parts.

    Turn the master fader down, or set your limiter's output so the integrated reading lands between −14 and −13 LUFS. Nothing else changes.

  2. 2Set the limiter ceiling to −1 dBTP.

    True peak is -0.5 dBTP. Lossy encoding (AAC/Opus) adds overshoot above this, and −1 dBTP is the headroom YouTube and Apple recommend.

    In your limiter, set the output ceiling to −1.0 dBTP (true-peak mode on). Then re-check.

  3. 354 % of the mix's energy is below 300 Hz.

    A phone speaker reproduces almost nothing below ~300 Hz, so that part of the mix does not exist there — what remains is thinner and the voice-to-music balance shifts.

    High-pass the music bed around 80–100 Hz and pull its low shelf down 3–4 dB; keep the voice untouched. Audition on the phone-speaker model after.

It measures three things, and models seven devices.

Every number on the page is a standard measurement or is labelled an estimate. The tool never changes your file: it tells you what it found and how to fix it in your own editor.

1 Loudness, the way YouTube measures it.

ITU-R BS.1770 integrated loudness (LUFS), loudness range and true peak, checked against ffmpeg's independent meter in the test suite. YouTube normalises to −14 LUFS: louder uploads are turned down, quieter ones are not turned up. The page says by how much.

2

Seven device models. Each is a short chain of filters — the same description drives what you hear and what the table says. They are typical members of each class, not measurements of your friend's phone; the page labels them modelled.

3

Voice over bed, as an estimate. Energy in the voice band against everything else, second by second, so you can see where the music swallows the dialogue. It is labelled an estimate because it is one.

The first three minutes of any file are free, on every device.

Free

$0

  • The first 3 minutes of any file, all seven devices
  • Integrated loudness, true peak, range, the timeline, the suggestions
  • Nothing uploaded, no account

Avyth subscription

$19a month

  • The whole file, however long
  • Also unlocks Thumbnail Preflight, Invoice Splitter and the other subscription tools
  • One key, pasted once per device. Cancel from a link in the email.

Uploads, accuracy and file limits.

Does my video get uploaded?

No. The file is decoded by your browser's own audio decoder and every measurement runs on your device. Close the tab and it is gone. The only thing that ever leaves is an email address, if you type one into the subscription form.

How accurate are the device models?

They are models — typical frequency responses for each class of device, expressed as a few filters — not measurements of a specific product. They are good enough to show what disappears and where the balance shifts; they are not a substitute for listening on the real thing once. If you want your own devices modelled, a sine sweep played on the device and recorded on a phone is enough; email support@avyth.com.

What does YouTube actually do to loudness?

YouTube normalises playback to about −14 LUFS integrated. If your video is louder it is turned down by the difference, quiet parts included; if it is quieter it is left alone. The page shows the difference and where your quietest passage lands after it. Spotify and Apple Podcasts use −14 and −16 respectively.

Why can't it fix the audio for me?

Because you would not be able to check the fix before it shipped, and the tool would be deciding for you. Every suggestion names the number it comes from and how to do it in an editor; you listen, then you decide. A one-click fix with a mandatory before/after listen is on the list.

What are the limits?

One file up to 300 MB and about 40 minutes of stereo audio (a browser tab holds only so much decoded sound; longer files are checked as an exported audio track). Anything the browser can decode: MP4, MOV, WebM, MKV, MP3, WAV, AAC, FLAC, OGG. Some codecs inside video containers cannot be decoded in a browser; the tool says so and asks for the audio track instead. The free tier measures the first 3 minutes; the subscription measures the whole file.

Now check your own.

Drop a video and listen on the phone-speaker model before you upload.