Encoder lift
+0.86dB
Codec-safe mastering
Added by YouTube's encoder alone.
Your master is safe in the WAV. Nobody plays the WAV.
Every platform re-encodes your file, and the encoder moves the peaks. We measure the decoded file and lower the ceiling until what the listener receives is correct.
- In your WAV-1.02dBTPwhat you export
- After YouTube-0.26dBTPlibopus 128k
- After our pass-1.51dBTPceiling -1.98 dBTP
Four things, four measurements
Not features. Each is the figure it produced.
Tonal spread, 3 tracks
1.50→1.09dB
Album mode
Separate generations, one shared target.
Spectrum wall found
20.1kHz
Built for generator output
Graded, never boosted above.
Decorrelated bass in mono
8.8→0.0dB
The low end, handled separately
Phones are mono. Under 120 Hz goes centre. Not from this repository
Interactive
Every level, measured through every codec
Nothing is computed in your browser. Each figure is a real encode, decoded and measured.
headroomnone left
Master aimed at -16 LUFS
- Came out
- -16.00 LUFSon target
- Worst after codec
- -5.81 dBTPok TikTok
- Ours
- -5.81 dBTPok ceiling -6.55
Master aimed at -14 LUFS
- Came out
- -14.00 LUFSon target
- Worst after codec
- -3.69 dBTPok YouTube
- Ours
- -3.69 dBTPok ceiling -4.55
Master aimed at -12 LUFS
- Came out
- -12.00 LUFSon target
- Worst after codec
- -1.83 dBTPok YouTube
- Ours
- -1.83 dBTPok ceiling -2.55
Master aimed at -10 LUFS
- Came out
- -10.56 LUFS0.56 dB short
- Worst after codec
- -0.27 dBTPtight YouTube
- Ours
- -1.55 dBTPok ceiling -1.98
Master aimed at -8 LUFS
- Came out
- -10.07 LUFS2.07 dB short
- Worst after codec
- -0.26 dBTPtight YouTube
- Ours
- -1.51 dBTPok ceiling -1.98
Master aimed at -6 LUFS
- Came out
- -9.31 LUFS3.31 dB short
- Worst after codec
- -0.48 dBTPtight YouTube
- Ours
- -1.22 dBTPok ceiling -1.76
- ok inside the limit
- tight under full scale
- clips over 0 dBTP
Every platform at every level, with codec, lift and playback
Before and after
Both players are level-matched
Louder reads as better, so an unmatched A/B only proves which file was pushed harder. Both play at -13.87 LUFS.
Source, unmodified
Master, level-matched
- Source
- Target profile
- Result
Album mode: three tracks, one target
- Tonal spread, per band
- 1.50→1.09dB
- Worst single band
- 4.79→2.81dBat 400 Hz
- Loudness spread
- 1.02→0.00LUnormalisation alone does this
- Stereo width spread
- 0.044→0.000
- Before, as generated
- After album mode
8 profiles, one source, nothing else changed
All level-matched to -14 LUFS, so the profile is the only variable left.
The source, no profile
Modern pop
5.36 → 2.08dB
Hip-hop and trap
5.35 → 2.05dB
EDM and dance
5.11 → 1.85dB
Lo-fi and chill
3.83 → 1.71dB
Acoustic and singer-songwriter
5.38 → 2.25dB
Rock and band
5.26 → 2.06dB
Pink reference
4.83 → 1.95dB
No EQ, loudness only
0.37 → 0.03dB
Every curve: ±9 dB, 40 Hz to 16 kHz. What each profile does, per band
Public reference
Delivery targets, and how much to trust each one
1 of 6 publishes its own loudness figure. Checked 2026-08-07.
Against the alternatives
What the others say about themselves
- Claims compared
- 11×4read off their own sites
- Cells we could not verify
- 17of 44never scored for us
- Rows where they win
- 5not a footnote
The WAV download is free, because paywalling it is the problem
Paid tiers buy volume, never better audio.
Single track
$2per track
- Everything in Free, one track
- No sign-up, nothing to cancel
- All platform renders
Unlimited
$12per month
- Unlimited masters
- Album mode across a release
- The distributor package
- Your own reference as target
Opening prices. EU customers pay VAT on top. Terms.
The honest part
What this cannot do
Every competitor could write this list and none of them do.
We cannot bring back what the encoder threw away
Above the wall in your spectrum there is only noise to amplify. We print where the wall is and leave it alone.
We do not make stereo out of mono
Widening means phase tricks, and phase tricks collapse on a phone speaker. We leave it mono and say so.
We master, we do not mix
A buried vocal cannot be lifted without lifting everything at that frequency. Fix the balance where it lives.
We do not distribute
You get the WAV and a render per platform. Getting it onto Spotify is DistroKid's job.
We cannot reach a level the limiter cannot reach
At the loudest setting above, the measured result lands short. We print what came out, not what you asked for.
We cannot predict how far the encoder will move your peaks
It depends on the material, so the tool encodes and measures per track. Anyone quoting one number has not measured it.
We are not running yet, and that is the biggest one
The measurements are real; the service is not open. No track record, no integrations. If you need something that works this afternoon, buy from someone else.
Questions that come next
aimastersongs.com
Drop a track in and read the numbers yourself
Three masters a month are free, and the free ones keep the WAV. Two dollars a track after that, twelve a month for as many as you like.