What is a fake hi-res audio file?
A file that claims a high sample rate (96 kHz, 192 kHz) or bit depth (24-bit) but actually contains content from a lower-resolution source, typically upsampled from a 44.1 kHz CD master or, worse, a lossy MP3. The container says hi-res; the audio is not. This is widespread in some streaming catalogues.
How does the FLAC & hi-res checker work?
It is a spectrum analyzer for audio files: it runs an FFT (fast Fourier transform) on your file and plots the frequency content. A 96 kHz file can hold energy up to 48 kHz and a 192 kHz file up to 96 kHz, and a native recording often shows some of it. A file made from a 44.1 kHz source shows a sharp cutoff just under 22.05 kHz with nothing above it, whatever the container claims. The result names what the spectrum shows and how sharp the cutoff is, because a native recording filtered at 20 kHz in mastering can look like an upsample.
Can I hear the difference between real and fake hi-res?
Almost never directly. Human hearing tops out at 20 kHz and falls fast above 16 kHz with age. The case for hi-res is mostly about the master, not the format: real hi-res releases often come from better masters with more dynamic range. The spectrum shows whether the file holds the high-rate content you paid for.
Is FLAC better than MP3?
FLAC (Free Lossless Audio Codec) is a lossless format: it makes a file smaller without discarding anything, so it decodes to exactly the samples that went in. MP3 is lossy: the encoder throws away what a model of hearing predicts you will miss least, and at 320 kbps that includes everything above about 16 to 19 kHz, the cutoff this checker shows. So FLAC is the better copy to keep, edit or convert from, and it takes more space: CD-quality FLAC usually runs 700 to 1,000 kbps against 320 for the best MP3. Whether you can hear the difference on your own system is a separate question, and a blind ABX test of the two files answers it.
Is my FLAC actually lossless, or transcoded from an MP3?
The same spectrum test answers that one. A lossy encoder discards the top of the band and lowpasses what is left, so an MP3 at 320 kbps stops dead around 16 to 19 kHz and AAC around 19 to 20 kHz. Re-encode that into FLAC and everything from then on is lossless, but the missing top never comes back and the brick wall sits exactly where the encoder left it. A FLAC ripped from a CD runs on to the 22 kHz Red Book wall instead. Drop the file in and read where the energy stops.
Anything your browser can decode: FLAC, WAV, AIFF, ALAC, MP3, AAC and OGG, plus DSD (DSF and DFF files), which the checker converts to PCM itself. The analysis runs on the decoded audio rather than the container, which is the whole point - the file extension, the sample rate in the header and the metadata are precisely the things a fake gets right.
Does the file get uploaded to your server?
No. The FFT analysis happens entirely in your browser using the Web Audio API. The file is decoded into an AudioBuffer locally, analyzed, and discarded when you close the page. We never see your audio files.
Does the checker work on a phone?
Yes, in the phone's browser, with nothing to download: Chrome on Android or Safari on an iPhone. Tap the File A box and pick the track from your downloads or the Files app. The decode and the spectrum run on the phone itself, as they do on a computer, and the file stays on it. A long 192 kHz track takes a few seconds longer, because the whole file is decoded into memory before the spectrum is drawn. To keep the checker on your home screen as an app, tap Install app under the checker on Android, or open Chrome's menu and choose Add to Home screen. On an iPhone, tap Share in Safari, then Add to Home Screen. The app opens straight into the checker, without the rest of the site around it, and once the page has loaded online it also opens and checks files with no connection.
Does it verify a FLAC file's integrity?
No, and the two checks answer different questions. A FLAC file stores an MD5 checksum of its audio, and the reference FLAC encoder's test mode (flac ‑t) decodes the file and compares the result against it, which tells you the file is intact and undamaged. It cannot tell you where the audio came from: an MP3 re-encoded to FLAC passes it perfectly. This checker answers that second question from the spectrum. Use both on a download you are unsure of.
Does it show the bitrate or bit depth?
Yes. It shows the sample rate and bit depth the file declares, its channels, length and size, and the average bitrate those give: the size in bits divided by the seconds, so a 30.0 MB file that runs 180.0 s averages about 1,400 kbps. Embedded cover art counts toward the size, so a heavily tagged file reads a little high. The header is exactly what a fake gets right, so the declared figures prove nothing on their own: the spectrum shows what the audio holds, and for a 24-bit file the bit-depth test below shows whether the extra bits are used.
How can I tell if a 24-bit FLAC is really 24-bit?
Drop it in. For a WAV, FLAC or AIFF file that declares 24 bits, the checker reads the samples themselves as well as the header. A 24-bit file made from a 16-bit master by padding has its lowest 8 bits at zero in every sample, and the result then says it holds 16-bit audio padded to 24 bits: the padding adds size and no detail. A file whose low bits carry signal is not padded, though that alone does not prove a 24-bit master, because a 16-bit recording with gain or dither applied after the conversion fills those bits too.
Can I check a whole album at once?
Yes. Pick several files in the File A box, or drop an album folder on it from your desktop, and the checker reads every audio file in it one after another, up to 100, skipping cover art and text files. A table lists each file's format, where its content reaches, the reading, the bit depth, and its loudness, true peak and PLR under a summary line, and Copy results as text puts it on the clipboard for a forum post. The chart shows one file at a time, and each row's Show button swaps it in. Nothing is uploaded, so a long 192 kHz album only takes longer.
Can it check DSD files (DSF and DFF)?
Yes, DSD64 to DSD512, in DSF or DFF. No browser decodes DSD, so the checker converts the file to PCM at 88.2 kHz itself, flat to 30 kHz, the way a player does when it plays DSD over a PCM input. DSD's own noise rises steeply above the audio band, so the reading looks for the one thing a DSD spectrum shows clearly: the wall a 44.1 or 48 kHz master leaves near 22 or 24 kHz, below the noise. With that wall, the DSD was most likely converted from a CD-rate or 48 kHz PCM master. Without it, the music runs on into the noise, which fits a native DSD recording or a high-rate master such as DXD. DST-compressed DFF files cannot be read, and of a stereo file longer than about eight minutes only the start is converted.
Can it measure a song's loudness in LUFS?
Yes. Every file gets its integrated loudness in LUFS, measured to ITU-R BS.1770-4, the standard streaming services and broadcasters level by; its true peak in dBTP, the waveform between the samples, interpolated four times below 96 kHz; its peak to loudness ratio (PLR), the true peak minus the loudness; and its loudness range (LRA) to EBU Tech 3342. In the album check each track gets its own figures, and the summary gives the loudness of the whole album measured as one program. A harder-mastered version of the same album shows a higher loudness and a lower PLR beside the original, which is how a squashed remaster shows up.
How much will Spotify turn my track down?
The checker works it out from the rule Spotify publishes. Spotify adjusts tracks to −14 LUFS, measured to ITU 1770, and Premium listeners can pick −11 LUFS (Loud) or −19 LUFS (Quiet) instead. A louder track is turned down by the difference, so a master at −8 LUFS plays 6 dB quieter at the standard level. A softer one is turned up only until its true peak sits 1 dB under full scale, except at Loud, which runs a limiter. An album played as an album is leveled as a whole, so its tracks keep their differences. Apple Music's Sound Check levels tracks too, but Apple's mastering guide gives no target figure, so the checker shows Spotify's.
