Supported file formats, sample rates and file sizes
Last updated 9 min readTags:AnalysisExporting
Which audio files Sonalix accepts, the best export settings for an accurate mix report, and how MP3s and 96 kHz files change the numbers.
Upload the same master twice, once as a WAV and once as a 128 kbps MP3, and the MP3's High band reads 19.3 dB lower. Both readings are correct. Each one describes the file Sonalix was given.
This page is for anyone about to upload a mix, and for engineers who want to know how Sonalix reads a file before trusting its numbers. Sonalix's supported file formats cover almost anything you are likely to export, in mono or stereo, but the report can only measure what is in the file. A lossless stereo export at 44.1 or 48 kHz gives the readings the analysis is built around. Lossy encodes and high sample rates shift specific numbers by measurable amounts, and the sections below show which numbers and by how much.
The export settings that give the most accurate report
If you only read one part of this page, read this. Export your finished mix or master with these settings, which are the same ones you would use for a distributor:
- Format: WAV, AIFF or FLAC
- Bit depth: 24-bit (32-bit float also works, see Phase 3)
- Sample rate: 44.1 kHz or 48 kHz
- Channels: stereo
- Processing: the mix as it stands, with no extra limiter or normaliser added just for the upload
Sonalix accepts WAV, AIFF, FLAC, MP3, M4A and OGG files in mono or stereo. For the most accurate report, upload a 24-bit stereo WAV, AIFF or FLAC at 44.1 or 48 kHz, exported before any MP3 or AAC conversion.
Four decisions that shape every upload
Every export dialog asks the same four questions: format, sample rate, bit depth and channels. Each one changes particular readings in your report, and each leaves a recognisable signature you can check afterwards.
Work through them in order. Format comes first because nothing later can restore what a lossy encoder has already removed. Sample rate comes next because it changes the frequency-based readings. Bit depth and channels mostly decide what certain readings mean, not whether they are accurate.
Phase 1: choose from the supported file formats, and prefer lossless
The upload panel lists six formats. Four are read exactly as they are. MP3 and M4A are converted on our server to a 16-bit WAV first, and then analysed.
| Format | How Sonalix reads it | What to expect |
|---|---|---|
| WAV (.wav) | Directly | Recommended. 16-bit, 24-bit and 32-bit float all work |
| AIFF (.aiff,.aif) | Directly | Identical results to the same audio as WAV |
| FLAC (.flac) | Directly | Lossless, identical results, smaller upload |
| OGG Vorbis (.ogg) | Directly | Lossy: the encoder's changes show in the report |
| MP3 (.mp3) | Converted to 16-bit WAV | Lossy: the encoder's changes show in the report |
| M4A (.m4a) | Converted to 16-bit WAV | Lossy if AAC; Apple Lossless is lossless but still converted to 16-bit |
Lossless formats store every sample of your export. In our tests, the same master as WAV, AIFF and FLAC produced identical results on every metric. Lossy formats (MP3, AAC, OGG) save space by discarding data, mostly at the top of the spectrum, and Sonalix measures what is left.
The size of the effect depends heavily on bitrate. A 128 kbps MP3 of our test master read the High band 19.3 dB lower than the WAV, the spectral centroid at 4,624 Hz instead of 6,231 Hz, and integrated loudness 0.43 LU quieter. A 256 kbps AAC of the same master read the High band only 0.3 dB lower. Those are figures from one test signal, not a rule, but the direction is consistent: the lower the bitrate, the duller and slightly quieter your mix will look.
True peak moves too, in either direction. From a master peaking at -0.01 dBFS, the OGG read +0.24 dBTP and the MP3 read -0.28 dBTP. The 16-bit conversion step also means an MP3 or M4A can never show a reading above 0 dBTP, even if the decoded audio went over.
Can I upload an MP3 to Sonalix? Yes. MP3s are analysed, but the High band, spectral centroid, true peak and loudness readings reflect the encoder as well as your mix.
A 256 kbps or higher file is fine for a rough check. For a report you plan to act on, upload the lossless export. Never upload an MP3 that has been opened in a DAW and exported again: it carries two rounds of encoding loss.
Phase 2: export at 44.1 or 48 kHz
Sonalix analyses your file at its own sample rate and does not resample it. The sample rate sets the highest frequency a file can hold, which is half the rate (the Nyquist limit): 22.05 kHz for a 44.1 kHz file, 24 kHz for 48 kHz.
44.1 kHz and 48 kHz give near-identical results, so use whichever your session runs at. Higher rates are where readings start to move.
We exported the same audio at 48 kHz and 96 kHz. Loudness, true peak and stereo correlation were unchanged, but sub stereo width read 0.084 at 96 kHz against 0.012 at 48 kHz. Every frequency band read about 3 dB higher, and the Sub band 5.8 dB higher. The severity thresholds stay the same whatever the rate, so the same mix can be rated differently at 96 kHz.
The decision is simple: if you work at 88.2, 96 kHz or above, export a 48 kHz copy for analysis. It is a standard delivery rate, and the copy takes seconds to export.
Low sample rates cause the opposite problem. The five bands and the frequency curve cover 20 Hz to 20 kHz. A 32 kHz file holds nothing above 16 kHz and a 22.05 kHz file nothing above about 11 kHz, so the High band and the top of the curve read as rolled off: missing content, not a dull mix. Below about 16 kHz the analysis fails altogether, because the High band starts at 8 kHz and the file holds nothing there.
What sample rate should I export my mix at for analysis? 44.1 kHz or 48 kHz. Both give the same results, and both are standard delivery rates.
Phase 3: pick a bit depth, and know what 32-bit float reveals
Bit depth makes no measurable difference to the report. The same audio exported at 16-bit and 24-bit produced identical readings in our test. Export 24-bit because it is the better habit for anything that may be processed again, not because the analysis needs it.
32-bit float is the exception worth understanding. Integer files (16 and 24-bit) cannot store anything above 0 dBFS, but float files can. A float WAV with peaks at +3.5 dBFS read +3.52 dBTP in Sonalix. That reading is accurate for the float file, but those overs will clip the moment the file is converted to 16 or 24-bit for distribution.
Two rules follow. A positive dBTP reading on a float file means you need to bring the level down before delivery. A reading of exactly 0.0 dBTP on an integer file means the file has probably already clipped. Sonalix measures the highest sample value without oversampling, as explained on the true peak page.
Does Sonalix support 32-bit float WAV? Yes, and it is the only format where the report can show overs above 0 dBTP.
Phase 4: check channels, length and file size
Stereo is the format the analysis is designed for. A mono file is accepted, but it is analysed as two identical channels, so stereo correlation reads +1.00 and sub stereo width reads 0. Those readings are correct for a mono file and tell you nothing, so ignore the stereo section of a mono report.
Surround and other multichannel files are not supported, and a file with more than five channels fails outright. Export a stereo fold-down. Upload the full mix rather than stems, because Sonalix analyses the finished stereo mix. Files shorter than about half a second also fail, so upload the whole track.
File size depends on the export settings. Uncompressed stereo WAV sizes work out as follows:
| Export | Per minute | 4-minute track |
|---|---|---|
| WAV 16-bit, 44.1 kHz | 10.6 MB | 42 MB |
| WAV 24-bit, 44.1 kHz | 15.9 MB | 64 MB |
| WAV 24-bit, 48 kHz | 17.3 MB | 69 MB |
| WAV 32-bit float, 48 kHz | 23.0 MB | 92 MB |
| WAV 24-bit, 96 kHz | 34.6 MB | 138 MB |
| WAV 24-bit, 192 kHz | 69.1 MB | 276 MB |
| MP3, 320 kbps | 2.4 MB | 9.6 MB |
If a WAV is slow to upload, export FLAC instead. It is lossless, gives identical results and is smaller, by an amount that depends on the material. Following Phase 2 also helps: a 24-bit, 48 kHz file is a quarter of the size of the same track at 192 kHz.
Can I analyse a mono file? Yes. It is measured as if the same signal played from both speakers, so the stereo readings don't apply.
How to spot an upload problem in your report
Each export decision leaves a signature. If a report looks wrong, check for these before changing your mix:
- High band far below Presence and a low spectral centroid on a mix you know is bright. You probably uploaded a low-bitrate lossy file. Re-upload the WAV, AIFF or FLAC.
- Stereo correlation +1.00 and sub stereo width 0. The file is mono, or both channels are identical.
- True peak above 0 dBTP. A 32-bit float file with overs, or an OGG whose decoded peaks went over. Either way, there is less headroom than you think.
- True peak at exactly 0.0 dBTP on a 16 or 24-bit file. The export has probably clipped.
- High band and curve flat above 11 or 16 kHz. The file's sample rate is 22.05 or 32 kHz. Re-export at 44.1 or 48 kHz.
Why an upload fails
A few kinds of file cannot be analysed at all:
- More than five channels. Export a stereo fold-down.
- Shorter than about half a second. Upload the full track.
- Sample rate below about 16 kHz, such as an 8 kHz voice memo. Re-record or re-export at 44.1 or 48 kHz.
- A damaged file, or one that isn't audio. Re-export from your DAW and try again.
A failed upload does not use up an analysis, because only completed analyses are counted.
The outcome: one file, measured as you made it
A 24-bit stereo WAV, AIFF or FLAC at 44.1 or 48 kHz, exported before any lossy conversion, is the file whose report describes your mix and nothing else. Every other choice adds the encoder, the sample rate or the channel layout to what you are reading. Export that file once, keep it as your delivery master, and upload your mix for a report you can act on.
Related pages
Was this page helpful?