1Upload the audio files you want brought to a consistent level — a whole set at once is exactly the case this is for.
2Pick a target: about -14 LUFS for streaming platforms, -16 LUFS for podcasts, -23 LUFS for broadcast.
3Run the analysis; integrated loudness and true peak are measured before any gain is decided on.
4Download the normalized Audio file, with the dynamics inside each track left untouched.
Normalize Audio Loudness FAQ
Does the source format affect normalisation?
+
It does. this format is read and written by the same streaming pipeline as everything else we support. That determines whether the corrected audio can be written back exactly or has to be re-encoded.
Anything specific to Audio here?
+
Yes — the codec is detected from the stream, so a mislabelled extension does not derail the job. Worth knowing when the goal is a consistent set rather than one file.
What loudness target should I aim for?
+
About -14 LUFS is what the major streaming platforms normalise to, -16 LUFS is the usual podcast target, and -23 LUFS is the broadcast standard. Going louder than the platform target gains you nothing — the platform simply turns it back down.
How does Normalize Audio Loudness decide the right level?
+
Concretely, the file is analysed for integrated loudness and true peak, then a single gain is applied to reach the target — no compression, no limiting, just the level. The file is measured first and the correction follows from the measurement, rather than a fixed gain being applied and hoped for.
Can I run Normalize Audio Loudness on several audio files at once?
+
Yes. The set queues in parallel with identical settings, which is how you process an album or a podcast back catalogue without repeating yourself.
Do tags and cover art survive?
+
Yes — ID3, Vorbis comments and MP4 atoms are carried across along with embedded artwork, so processed files still sort correctly in a music library.
Is there a file size limit on Normalize Audio Loudness?
+
Yes: free accounts process audio up to 15 MB per file whichever codec you brought; ffmpeg handles the decode and encode. In practice that limit only bites on uncompressed WAV; a compressed file of the same length is nowhere near it.
Will Normalize Audio Loudness lower the quality of my audio files?
+
It depends on the source. Lossless input (WAV, FLAC, AIFF) is edited sample-exactly and loses nothing. Lossy input has already discarded detail once, so where the job can be done by copying the encoded stream rather than re-encoding it, that is the path taken.
Why does a Word site host Normalize Audio Loudness?
+
WORD.to is built around the editable end of a document's life — the DOCX that is still being written, still being styled and still being argued over, before anyone flattens it for sending. Office files are containers full of other people's media — images, embedded audio, fonts — so the work people need on them is usually the work they would need on those contents anyway. Normalize Audio Loudness shares the upload, the caps and the account with the conversions for that reason.
What should I do with the result once Normalize Audio Loudness is finished?
+
The converter on this site takes documents out to PDF for sending, to images for embedding and to plain text for anything that has to read them programmatically, and back the other way. Doing that afterwards keeps the editable original around, which is the part you cannot get back once it has been flattened.
Is Normalize Audio Loudness here the same tool the sibling sites run?
+
The engines are shared — the same document toolchain, the same workers, the same limits. What a Word site adds is a view on what survives leaving the Office format and what does not, which is the question every one of these jobs actually turns on. It also starts from one fact about the format this site is named after: the document is a zip of XML, so text edits are cheap and the weight is almost always the embedded images.
Do I need an account, and does anything get kept?
+
No account, and nothing is kept: uploads are deleted from the workers shortly after the job finishes, nothing is read and nothing is indexed. Free accounts exist for history and batch size, not for access.
