Getting started

Quick start

1. Open LoudSense at squaresmile.de/app/loudsense or from the dashboard. Checking files needs no account.
2. Drop a WAV, MP3, AAC, FLAC or OGG onto the drop zone, or click Choose files or Choose a folder. On a phone, tap the drop zone to choose a file. With the keyboard, Tab to the drop zone and press Enter or Space. Up to 1 GB per file, up to a hundred files at once.
3. Watch the progress bar. A 20-minute episode takes a few seconds on a laptop.
4. Pick a target group — Podcast, Audiobook, Music streaming, Broadcast or Live concert — and read the verdict rows. Each one says OK, too loud, too quiet, peaks too hot or clipped, and what to change. On a phone the verdict is the first thing under the file name.
5. Press Fix to this on a row to download a fixed WAV for that target, or read the panel first: waveform, marks, curve, and the stats row.
6. Copy as text puts the numbers and verdicts on the clipboard as text, ready for a message to the client or the band.

The panel

• Waveform — the whole file, both channels folded into one lane.
• Red lines — clips: runs of full-scale samples. One line per run.
• Cyan marks at the top — the ten loudest peaks.
• Orange curve — momentary loudness (LUFS-M, 400 ms window) over time, mapped from −60 LUFS at the bottom to 0 at the top. It shows where an episode drifts and where the chorus lands.
• Yellow line — the playhead. Click anywhere in the waveform to seek; the transport has play, pause, stop, loop and volume.
• Time ruler underneath.

The stats row

Every number is explained on Reading the numbers. The line under the row carries format, sample rate, channel layout (mono, stereo, 5.1), length, and two more numbers for audiobook specs: RMS and noise floor. A 5.1 file is measured on every channel per BS.1770, the LFE left out.

Health

Under the stats, LoudSense lists what it found beyond loudness: channel balance and peaks, stereo correlation (mono compatibility), DC offset, silence at the start and end, silent gaps longer than three seconds, digital dropouts (runs of exact zeros), and the clip count. A clean file gets one green line. These are the editing mistakes that reach a platform unnoticed; each finding says what to do.

Selecting a range

Drag across the waveform to select a part of the file — the intro, an ad read, the chorus. A yellow box shows loudness, peak, LRA and clips for that range alone, and Loop selection plays it round. Click anywhere to seek; Esc clears the selection; Space plays and pauses; ← → skip five seconds.

Fix it

Press Fix to this on a verdict row, or pick a target under the verdicts — the menu starts on the group you are looking at — choose 24- or 16-bit (16-bit is dithered), and Download fixed WAV. LoudSense applies the gain the verdict asked for and, when the raised peaks would cross the target's ceiling, a true-peak limiter at that ceiling. Then it measures the result and corrects it until it lands within 0.1 LU of the target with the true peak under the ceiling — heavy limiting no longer leaves the file short. The toast reports the LUFS-I and true peak it reached.

If you will deliver MP3 or AAC, say so in the menu: the encoder adds overshoot, so Fix it lowers the ceiling and checks the encoded result until it holds. A 5.1 file keeps all its channels.

The original is untouched; the fixed file downloads and is analysed as a new row, so you see before and after side by side. It is a normaliser, not a mastering chain: no EQ, no compression, no colour. A clipped file stays clipped — the cut-off peaks cannot be restored, so fix the master before the clip.

Encode preview

Choose a bitrate and Encode & measure: the file is encoded to MP3 in your browser and the result appears as a new row. Compare its true peak and clips with the original — that is what the encoder will do to your peaks before you upload anything. The preview is a real MP3 you can download from the list if you want it.

In Chrome and Edge the codec menu also offers AAC: the file is encoded with the browser's own AAC encoder and measured straight away. AAC previews are numbers only — no download, no playback — because the browser encodes but does not write an .m4a container.

Comparing files

Drop two or more files and a Compare table appears under the panel, one row per file, with a CSV download. The classic use: the WAV master and the MP3 you are about to upload. If the MP3 shows clips the WAV does not, the encoder added overshoot — see why the MP3 clips.

• Reference. Press ★ on a file to make it the reference; every other file then shows its delta ("2.1 LU quieter, LRA +4, true peak +0.3").
• Album. With two or more original files loaded, the album loudness appears — the number Spotify's and Apple's album normalisation act on — with the loudest and quietest track and what each service does to the whole album.
• Fix & export holds Fix it, the encode preview and the report: Export PNG saves the waveform, stats and verdicts as one image for a message to the client or the band; Copy as text does the same as text.

Speech or music?

Under the panel a thin strip marks where the file is speech (blue), music (orange) and silence, with the share of each and the loudness of just the speech and just the music. A podcast with a music bed reads one LUFS-I for the whole episode; the split tells you whether the voice itself sits where you want it, and by how much the intro jingle sticks out. The classification is a heuristic (modulation, spectral tilt, zero crossings) — treat it as a guide, not a transcript.

Episode history

Every original file you analyse is remembered in this browser — the numbers only, never the audio. When a file from the same series comes back (same name minus the episode number or date), the History tab shows a badge and a sparkline of LUFS-I over the last twelve, the spread, and a table. More than 2 LU between episodes means listeners will reach for the volume. The history holds the file name, the numbers and the date. Clear it any time; nothing of it is sent to the server.

A/B and "hear it normalised"

With two files loaded, the A/B tab plays them in sync and lets you flip between them. Level-matched turns the louder one down to the quieter one's loudness so you compare sound, not volume — the trick every mastering engineer uses. Hear A as plays the current file at a platform's level: pick Spotify and you hear what the listener hears after normalisation — turned down when it is too loud, turned up as far as Spotify would. Playback runs through the browser's audio engine, so the level change works on iPhone and iPad too.

Charts

The Charts tab shows a histogram of momentary loudness (a tall, narrow hill is a consistent file; two hills are a quiet and a loud part) and the short-term curve over time with the LRA band shaded and LUFS-I dashed. Download the curve as CSV gives a row per 100 ms for a spreadsheet or an editor marker list.

Folders, phones and sharing

• Choose a folder (or drop one) analyses every audio file in it — an album, a playlist, a season — up to 100 at a time, in name order. The list turns compact, the Compare table and the album loudness follow.
• On Android with the app installed (Chrome), LoudSense appears in the system share sheet; iPhone and iPad do not support this — open LoudSense and choose the file instead. Share a file from the recorder or Files, and LoudSense opens with the file waiting: tap Analyse to measure it. Very large files are refused with the size limit. The file is handed from the sharing app to the page inside the device. If you are offline or not signed in yet, it waits in this browser until you open LoudSense again.
• Share… sends the PNG panel through the share sheet (WhatsApp, Mail, AirDrop) where the browser supports it.

Which files work where

• WAV (including RF64 and 5.1) streams in every browser — memory stays flat however long the file.
• MP3 streams in browsers whose WebCodecs can decode MP3 (Chrome and Edge, for example).
• AAC, FLAC, OGG, AIFF, and MP3 where streaming is not available, go through the browser's own decoder in one piece, so they are limited to 250 MB on a laptop and 60 MB on a phone. They are measured at the file's own sample rate. For a long file in one of these formats, export a WAV.