Free tool · loudnorm · Runs in your browser
Normalize audio loudness
Short answer
The tool runs FFmpeg's loudnorm filter with an integrated loudness of -16 LUFS, a true peak limit of -1.5 dBTP and a loudness range of 11 LU, then saves an MP3. On three versions of the same speech recording, measured at -23.0, -43.0 and -13.3 LUFS, the results were -16.0, -16.0, -15.9 LUFS. Apple Podcasts recommends about -16 LKFS. Nothing is uploaded.
Maximum file size: 100 MB. The first run loads a ~30 MB engine, once per visit. Nothing is uploaded - processing runs on your device.
Key takeaways
- The target was hit on all three recordings: -23.0 became -16.0 LUFS, -43.0 became -16.0 LUFS, -13.3 became -15.9 LUFS. The largest miss was 0.1 LU.
- That includes a recording 27 LU too quiet. It was brought up to -16 LUFS, which also raises the background by the same amount, so a noisy quiet recording becomes a loud noisy one.
- True peak stayed near the limit: -1.4 dBTP, -1.4 dBTP, -2.2 dBTP against a -1.5 dBTP target. 2 of 3 ended 0.1 dB above it after the MP3 encode, still under Apple's -1 dBFS ceiling.
- Loudness range barely moved on speech: 2.5 LU before and 2.7 LU after for the recording as published. We did not test music, where a dynamic mode can change the dynamics more.
- The output is an MP3 at LAME quality 2 (108 kbps, 112 kbps, 108 kbps for our mono speech), at 48 kHz. Files up to 100 MB are accepted.
What the tool does to your file
FFmpeg describes loudnorm as EBU R128 loudness normalization with single-pass and double-pass modes, and says that in dynamic mode the sound is upsampled to 192 kHz to detect true peaks, so the output rate should be set explicitly. The tool uses a single pass, which is the dynamic mode. It does not measure the file first and does not use the linear mode, which needs measured values.
| Setting | Value | What it means |
|---|---|---|
| Integrated loudness | I = -16 LUFS | The average loudness of the whole file |
| True peak | TP = -1.5 dBTP | The highest peak after oversampling |
| Loudness range | LRA = 11 LU | How much loudness may vary within the file |
| Mode | One pass, dynamic | No pre-measurement of the file |
| Output | MP3, LAME quality 2 | About 100 to 115 kbps for mono speech |
The filter string is loudnorm=I=-16:TP=-1.5:LRA=11. FFmpeg's defaults are I = -24, TP = -2 and LRA = 7; the tool changes all three.
Measured: three recordings, one target
Measured on 30 seconds of a LibriVox reading of The Heart of a Mystery (public domain), as published (128 kbps mono MP3) and as a lossless WAV decode of the same recording. The tool's own ffmpeg arguments ran on the same single-thread WebAssembly engine your browser downloads; times are from a headless run and your device will differ.
Loudness and true peak were measured with ffmpeg's ebur128 filter before and after. The second and third rows are the first recording made 20 dB quieter and 10 dB louder.
| Recording | Before: loudness, range, true peak | After: loudness, range, true peak | Miss against -16 LUFS | Output |
|---|---|---|---|---|
| speech as recorded | -23.0 LUFS, 2.5 LU, -6.1 dBTP | -16.0 LUFS, 2.7 LU, -1.4 dBTP | 0.0 LU | 407,153 bytes, 108 kbps, 48 kHz |
| very quiet speech (20 dB lower) | -43.0 LUFS, 2.5 LU, -26.1 dBTP | -16.0 LUFS, 2.7 LU, -1.4 dBTP | 0.0 LU | 419,705 bytes, 112 kbps, 48 kHz |
| loud speech (10 dB higher, limited) | -13.3 LUFS, 2.9 LU, 0.0 dBTP | -15.9 LUFS, 2.8 LU, -2.2 dBTP | +0.1 LU | 405,329 bytes, 108 kbps, 48 kHz |
The Normalize audio tool on three versions of a 30 second speech recording.
Is -16 LUFS the right target?
It depends where the sound is going. Apple's audio requirements recommend preconditioning podcast audio so the overall loudness is around -16 dB LKFS, with a tolerance of one dB, and a true peak no higher than -1 dB FS. The tool's target and its measured results sit inside that band: -16.0, -16.0, -15.9 LUFS.
Spotify uses a different target for music: it adjusts tracks to -14 dB LUFS during playback, measured to the ITU 1770 standard, and its help page says the web player and third-party devices do not use loudness normalization. A file normalized to -16 is two LU below that figure.
| Service | What it says | This tool |
|---|---|---|
| Apple Podcasts | Around -16 dB LKFS, plus or minus 1 dB; true peak not above -1 dB FS | Target -16 LUFS; measured -16.0, -16.0, -15.9 |
| Spotify | Tracks are adjusted to -14 dB LUFS during playback; the web player does not normalize | Two LU quieter than -14 |
Those are two services' statements, not a rule for all platforms. Check the requirements of the place you are publishing to.
What a one-pass normalizer cannot promise
- An exact result. We measured misses of up to 0.1 LU on speech; other material may differ more.
- The true peak limit after the MP3 encode. Two of our three outputs measured -1.4 dBTP against a -1.5 target, because lossy encoding adds a little peak.
- A quiet recording without its noise. Gain raises everything, noise included; clean the sound first if it is noisy.
- Music mastering. Pop music is often mastered much louder than -16, and a dynamic mode would reduce its level and may change its dynamics. We did not test music.
Free tools:Remove background noise from audioRemove silence from audioCut an audio fileWhat is LUFS?
Order of operations for a clean voice recording
- Remove background noise first, so gain does not lift it.
- Remove long silences if you want a tighter file.
- Normalize last, so the final level is the one you measured.
Questions people ask
How do I normalize audio to -16 LUFS?
Choose the file and run the tool. It applies loudnorm with I = -16, TP = -1.5 and LRA = 11 and saves an MP3. Our test recordings landed at -16.0, -16.0, -15.9 LUFS.
What is the difference between normalizing and compressing?
Normalizing here sets the overall loudness of the file. In dynamic mode loudnorm can also change the dynamics a little: speech loudness range moved from 2.5 to 2.7 LU in our test.
Why not -14 LUFS?
That is the figure Spotify says it adjusts music to on playback. The tool uses -16, the level Apple recommends for podcasts.
Does it work on very quiet recordings?
Yes: a recording at -43.0 LUFS came out at -16.0. The background noise is raised by the same amount.
Does it lower the sound of a loud recording?
Yes. A recording at -13.3 LUFS with peaks at 0.0 dBTP came out at -15.9 LUFS and -2.2 dBTP.
What format do I get back?
An MP3 at LAME quality 2 and 48 kHz. Files up to 100 MB are accepted, and nothing is uploaded.
Related
Sources and how we checked
- FFmpeg Filters Documentation (read 2026-10-06)
- Apple Podcasts for Creators: Audio requirements (read 2026-10-09)
- Spotify Support: Loudness normalization (read 2026-10-06)
Reviewed October 9, 2026. The measurements ran the tool’s own ffmpeg arguments on the same single-thread WebAssembly engine your browser downloads. The script, the results and the saved documentation are in data/seo/hub/evidence/2026-10-09/tools.