Troubleshooting

Symptom, likely cause, fix — for files that will not analyse, numbers that look wrong, and playback that stays silent.

Symptom	Likely cause & fix
"Files up to 1 GB — "name" is … GB."	The file is over the cap. Export a lower bit depth or a shorter section; a 24-bit 48 kHz stereo WAV is about 16 MB per minute.
"This browser cannot stream-decode … the whole file must be decoded at once"	The format is not WAV, or it is an MP3 and this browser's WebCodecs cannot decode MP3. Export a WAV, use Chrome or Edge, or trim the file under the size shown.
The bar sits at 99 % for a long time	For MP3 and formats that decode whole, the bar follows the measurement, which runs at roughly 50–100× realtime: a one-hour file takes about a minute on a laptop. If it never finishes, reload once (a stale service-worker build) and write to support with the format.
"Could not decode this file"	The browser does not understand the codec (some AAC profiles, DSD, video containers), or the file is truncated. Export a WAV from the source.
"No MPEG audio frames found"	The file has an .mp3 name but is not MPEG audio and the browser's own decoder could not read it either. Re-export it as WAV or MP3.
LUFS-I reads about 3 LU lower than expected	The file is mono. Loudness sums channels; the same programme as dual-mono stereo reads 3 LU louder. For a podcast, export stereo and aim for −16 — see Mono or stereo?.
A 5.1 file reads differently from a stereo downmix	Correct: a 5.1 file is measured on every channel with the BS.1770 weights (surrounds +1.5 dB, LFE left out). A downmix is a different signal.
True peak is higher than Peak	Normal: the waveform between samples overshoots. True peak is what platforms check.
Clips on the MP3 but not on the WAV	Encoder overshoot — at 128 kbps it can be 2 to 4 dB on a loud master. Leave more headroom, or run Fix it with MP3/AAC delivery so it lowers the ceiling.
The verdict says "Clipped"	The file has runs of full-scale samples. Turning it down will not bring the cut-off peaks back; go back to the master before the clip and limit it to the ceiling.
A broadcast row says "Out of spec — would be rejected"	Broadcasters do not normalise a file, they send it back. Fix it to the target, or follow the fix on the row.
The orange curve is flat at the bottom	The file is very quiet (below −60 LUFS momentary) or silent. Check the export's routing.
Playback is silent	Browser autoplay rules: click play again, and check the volume slider and the OS output. The waveform still draws without playback.
Playback works but the playhead does not move	The tab is in the background; browsers throttle animation. Bring it to the front.
Analysis is slow on a phone	Phones decode at a few times realtime. A 60-minute WAV can take a minute. WAV and MP3 stream, so memory is not the problem; time is.
Verdict says "too quiet" with a limiter in the fix	The gain needed would push true peak over the platform's ceiling. Raise the level with a true-peak limiter at that ceiling, not with plain gain — or press Fix it.
"Fix it writes a … MB WAV — the limit is 400 MB"	The fixed file would be too large to hold in memory. Choose 16-bit, or fix a shorter export.
The fixed file reads slightly off target	Fix it re-measures and corrects until it lands within 0.1 LU, so a larger gap means the file needs more limiting than is sensible — the toast says so. A fixed file that is then encoded to MP3 or AAC also reads differently: compare the encoded file.
Encode preview takes long	Encoding runs at roughly 30–60× realtime in the browser; a one-hour episode takes about a minute. Keep the tab in front.
Noise floor shows a number on a clean master	The estimate takes the quietest 100 ms blocks that are not digital silence — a fade-out or a breath counts. It matters only against the audiobook spec.
"AAC preview needs a browser with WebCodecs audio encoding"	Your browser has no AAC encoder (Chrome and Edge do). Use MP3 preview elsewhere; the encoder behaviour on peaks is similar.
The speech/music strip is wrong for my file	It is a heuristic on modulation and spectral tilt. Sung vocals, spoken word over drones and heavily compressed speech confuse it. The per-class numbers are still right for what it labelled; the whole-file LUFS-I is unaffected.
The History tab shows no trend	It needs two files of the same series in this browser (same name minus the episode number or date), and only originals count — fixed, trimmed and preview rows are skipped.
A/B plays only one file	Pick a second file under "B" first; without it the button plays A alone with the "Hear A as" turn-down. Both files must have finished analysing.
My targets do not show on another device	Sign in on both; targets sync through the account. Without an account there are no own targets — checking files works without one.
The share sheet does not list LoudSense	Share targets work only in the installed app on Android (Chrome) and in Chrome or Edge on a computer. iPhone and iPad do not support them: open LoudSense and choose the file instead.
I shared a file but nothing is analysed	Shared files wait for a tap: press Analyse in LoudSense. Files above the size limit are refused. If you were offline or signed out, the file waits in this browser until you open LoudSense again.
I can't reach the file picker with the keyboard	`Tab` to the drop zone and press `Enter` or `Space`. Reload once after an update if it does not respond.
Health says "wide stereo image" on speech	The channels differ a lot; for a mono voice that usually means a stereo reverb or two mics. Fine for music, worth a check for a podcast.

Still stuck?

Copy the report (the button under the panel) and send it with the file format and your browser to support@squaresmile.de. The file itself is not needed for most questions, and you never have to send it.

Airchain

Still stuck?