Audio Loudness Normalizer

Source audio:

WAV, MP3, M4A, AAC, OGG, FLAC, or WebM; up to 12 MB and 60 seconds.

{{ sourceFile ? `${formatBytes(sourceFile.size)} · ready to normalize` : 'Drop one file or browse from your device' }}

Ignored {{ ignoredFileCount }} extra {{ ignoredFileCount === 1 ? 'file' : 'files' }}. This tool processes one file at a time.

Integrated loudness target:

Use the destination specification. These common values are starting points, not universal delivery rules.

Sample-peak ceiling:

Gain is reduced when necessary so decoded samples remain at or below this ceiling.

Output edge fade:

The neutral default is no fade. Use a short fade only when the source begins or ends abruptly.

Normalized audio

{{ row.label }}{{ row.display }}

The requested target would exceed the sample-peak ceiling, so the tool stopped at {{ formatLufs(computation.values.output_lufs) }}. Use a reviewed dynamics workflow if the exact target is mandatory.

Verified WAV

Preview and download use the same locally generated PCM16 file.

Revision {{ outputRevision }}

The chart renderer is unavailable. The same values remain in the audio and log tabs.

Signal	Value	Meaning	Copy
{{ row.label }}	{{ row.display }}	{{ row.detail }}

Two audio files can reach similar peaks yet sound very different in overall level. Integrated loudness measures a complete programme over time, giving quieter and louder passages a combined value in loudness units relative to full scale (LUFS). Peak level answers a different question: how close the largest digital sample comes to the maximum representable level.

Loudness normalization applies one constant gain to the whole file. That preserves the relationship between quiet and loud passages, unlike compression or automatic leveling. If the requested gain would push a sample above the selected ceiling, the gain must stop early and the programme will finish below its loudness target.

Audio loudness and peak terms
Term	Meaning	Interpretation limit
Integrated loudness	Gated average programme loudness expressed in LUFS.	It estimates perceived programme level, not the loudness of every moment.
Sample peak	The largest decoded sample amplitude expressed in decibels relative to full scale (dBFS).	It can miss an inter-sample peak that appears after digital-to-analogue conversion or lossy encoding.
Loudness target	The intended integrated level after gain.	A preset is a starting point; the destination specification remains authoritative.
Peak ceiling	The highest allowed decoded sample peak after gain.	A lower ceiling may prevent the target from being reached.

Targets depend on the destination. Speech, music distribution, broadcast, and an internal editing handoff may use different loudness and peak policies. A file normalized for one destination should not be assumed compliant with another.

Loudness normalization also cannot repair noise, distortion, clipping already present in the source, uneven dialogue, or excessive dynamics. Those problems need editing, limiting, compression, or a new mix rather than a different single gain value.

How to Use This Tool:

Use the delivery specification to choose the target and ceiling before processing the file.

Choose one WAV, MP3, M4A, AAC, OGG, FLAC, or WebM audio file up to 12 MB and 60 seconds. The decoded source must be mono or stereo and at least 400 milliseconds long.
Select Integrated loudness target. The available −14, −16, −18, and −23 LUFS values are common starting points rather than universal delivery rules.
Select the Sample-peak ceiling. Use more headroom when later lossy encoding or an external specification calls for it.
Leave Output edge fade at None unless an abrupt start or end needs a short 10, 25, or 50 millisecond linear fade. The fade does not change the measured gain decision.
Normalize the audio, then compare source loudness, applied gain, output loudness, and output sample peak. Preview the generated WAV before downloading it.

Interpreting Results:

When Peak limited is false, the applied gain equals the requested difference between source and target loudness. When it is true, the sample-peak ceiling allowed less gain, so the output loudness remains quieter than the selected target.

Positive gain raises the whole programme; negative gain lowers it.
An output sample peak at the ceiling is expected after a peak-limited run, but it is not a true-peak compliance result.
A target match describes integrated loudness only. Listen for distortion, clicks, noise, and objectionable dynamics before delivery.

Technical Details:

Measurement uses mono or stereo audio decoded at 48 kHz. Each channel passes through the two-stage K-weighting filter defined for loudness measurement. Mean-square energy is then measured in 400 millisecond blocks with 75 percent overlap before two-stage gating removes silence and very quiet passages from the integrated result.

Rule Core:

Integrated loudness gating rules
Pass	Rule	Purpose
Absolute gate	Keep blocks whose loudness is strictly greater than −70 LUFS.	Remove blocks below the absolute measurement floor.
Relative threshold	Measure the absolute-gated blocks, then subtract 10 dB.	Set a programme-relative boundary.
Final gate	Use the greater of −70 LUFS and the relative threshold, then keep blocks strictly above it.	Choose the blocks used for integrated loudness.
Peak guard	Applied gain is the smaller of requested loudness gain and available sample-peak headroom.	Keep decoded samples at or below the selected ceiling.

Formula Core:

The gated mean-square energy becomes integrated loudness. The requested gain is then limited by available peak headroom and converted to a linear amplitude multiplier.

L = - 0.691 + 10 \log_{10} (\bar{E})

G_{requested} = L_{target} - L_{input}

G_{applied} = \min (G_{requested}, P_{ceiling} - P_{input})

g = 10^{\frac{G_{applied}}{20}}

Here, L is loudness in LUFS, Ē is the mean energy of the final gated blocks, G is gain in decibels, P is sample peak in dBFS, and g is the linear amplitude multiplier. Output loudness equals input loudness plus applied gain; output sample peak equals input sample peak plus applied gain, capped at 0 dBFS.

Transformation Core:

The same constant multiplier is applied to every decoded sample. Values are limited to the digital range, the optional edge fade is applied after gain, and the result is encoded as a 16-bit PCM WAV. The generated WAV is decoded again to confirm channel count, duration, and the reviewed sample-peak ceiling before it becomes available.

No compressor or limiter changes the programme envelope. When constant gain cannot reach both the loudness target and peak ceiling, the peak ceiling wins and the result records that decision.

Accuracy and Privacy Notes:

The measurement follows the ITU-R BS.1770-5 gated loudness stages used here for 48 kHz mono or stereo material. The peak guard measures decoded samples rather than oversampled true peak, so it is not a true-peak certification method and cannot guarantee acceptance after another codec or platform processes the file. An optional edge fade is applied after the gain decision and is not remeasured for integrated loudness; it can make a short file finish slightly quieter than the projected output value.

The source audio, decoded samples, loudness blocks, and generated WAV remain in browser memory for the current page. They are not uploaded, logged, stored remotely, or placed in the shareable URL.

Worked Examples:

Target blocked by peak headroom

A programme measured at −20 LUFS with a −5 dBFS sample peak needs 6 dB of gain to reach −14 LUFS. A −1 dBFS ceiling leaves only 4 dB of headroom, so the applied gain is 4 dB. The result reaches −16 LUFS and −1 dBFS, with Peak limited shown because another 2 dB would cross the ceiling.

References:

Recommendation ITU-R BS.1770-5: Algorithms to measure audio programme loudness and true-peak audio level, International Telecommunication Union, November 2023.
EBU Tech 3341: Loudness Metering, European Broadcasting Union, November 2023.