The short answer
- Peaks: 5 of 5 LuvLang (Professional) files stayed within Spotify’s true-peak guidance (15 of 15 across all three LuvLang plans). LANDR: 0 of 5 (every file peaked at -0.1 dBTP). BandLab: 0 of 5, with 5 of 5 peaking above 0 dBTP, which means clipping.
- Loudness: LuvLang landed -14.3 to -13.9 LUFS, at Spotify’s -14 reference. LANDR -14.3 to -11.7; BandLab -12.2 to -8.9. Louder files are simply turned down by Spotify, Apple Music and YouTube.
- Tone: LANDR changed the tone the least (0.7 dB on average), BandLab 1.1 dB, LuvLang the most (1.4 dB). More change is not better or worse by itself; it depends on the mix.
The results
| Service | Loudness (LUFS) | True peak (dBTP) | Files within Spotify’s peak guidance | Tone change vs the mix |
|---|---|---|---|---|
| LuvLang (Professional) | -14.3 to -13.9 | -2.3 to -1.1 | 5 of 5 | 1.4 dB |
| LANDR | -14.3 to -11.7 | -0.1 | 0 of 5 | 0.7 dB |
| BandLab | -12.2 to -8.9 | 0.1 to 0.8 | 0 of 5 | 1.1 dB |
“Spotify’s peak guidance” is Spotify’s own advice for masters: keep true peak below -1 dB, and below -2 dB if the master is louder than -14 LUFS, to avoid extra distortion when the file is encoded. “Tone change” is the average size of the per-band change versus the original mix after matching loudness (six bands, 20 Hz to 16 kHz).
Every file, mix by mix
| Mix | LuvLang (Professional) | LANDR | BandLab |
|---|---|---|---|
| Mix 1 | -14.0 LUFS / -2.3 dBTP | -11.7 LUFS / -0.1 dBTP | -8.9 LUFS / 0.8 dBTP |
| Mix 2 | -14.2 LUFS / -2.1 dBTP | -13.8 LUFS / -0.1 dBTP | -11.8 LUFS / 0.3 dBTP |
| Mix 3 | -13.9 LUFS / -2.1 dBTP | -12.7 LUFS / -0.1 dBTP | -11.3 LUFS / 0.1 dBTP |
| Mix 4 | -14.3 LUFS / -2.1 dBTP | -14.3 LUFS / -0.1 dBTP | -12.2 LUFS / 0.1 dBTP |
| Mix 5 | -14.1 LUFS / -1.1 dBTP | -12.2 LUFS / -0.1 dBTP | -10.6 LUFS / 0.7 dBTP |
Two controls, for scale
To show what “no mastering” and “automatic reference matching” look like on the same mixes, we ran two free tools too: plain loudness normalization (ffmpeg loudnorm, -14 LUFS, -1 dBTP) and Matchering, an open-source reference-matching master.
| Service | Loudness (LUFS) | True peak (dBTP) | Files within Spotify’s peak guidance | Tone change vs the mix |
|---|---|---|---|---|
| Loudness-normalize only (control) | -14.3 to -13.7 | -4.3 to -1.0 | 4 of 5 | 0.0 dB |
| Matchering (open-source control) | -14.9 to -13.7 | -6.1 to -5.6 | 5 of 5 | 2.1 dB |
What the numbers show
- Peaks are where the services differ most. A file that peaks at -0.1 dBTP or above 0 can distort once a streaming service encodes it. LuvLang’s ceiling is set by the same rule Spotify publishes, including the stricter -2 dB ceiling for masters louder than -14 LUFS.
- Loud is not a win on streaming. BandLab’s files were the loudest (-12.2 to -8.9 LUFS); on Spotify they are turned down to about -14, so the extra loudness only costs dynamics: BandLab’s crest factor dropped 1.7 dB on average, versus 0.6 dB for LuvLang.
- LANDR is the most transparent on tone. If your mix is already where you want it, a lighter touch is a real advantage. LuvLang now corrects a well-balanced mix more gently on Professional and Studio, and lets you choose the tone with one-tap sounds and Fine-tune instead.
How we measured
Five 60-second excerpts of unmastered mixes from one live band (pop and rock), each sent through every service. BandLab masters were made on BandLab’s website on November 4, 2026 with its default preset at default strength. LANDR masters were made on LANDR’s website before our test; the settings used for them were not recorded, so LANDR’s loudness in particular may reflect a setting choice. LuvLang masters were rendered on November 18, 2026 from the live site, Professional plan, with default settings. Loudness and true peak were measured with ffmpeg’s EBU R128 meter (ITU-R BS.1770-4, 4× oversampled true peak). Tone change compares each master with its source after matching loudness.
Limits. Five excerpts from one band is a small sample; other mixes and genres can behave differently, and each service may have changed since we made these files. We make LuvLang and tested our own service alongside the others; we report only measurements, not listening preferences, and we have not published anything we did not measure. Questions or a different result? hello@luvlang.studio.
Using this data? You’re welcome to cite it with a link to this page.
