Spectrogram Analyzer & Transcode Detector

Drop an audio file to see its spectrogram and check whether it's genuinely lossless or a fake — a re-encoded MP3 dressed up as FLAC, or upsampled "hi-res". Everything runs in your browser; nothing is uploaded.

100% Free Nothing Uploaded Works on Mobile

Drop an audio file here or click to select

MP3, FLAC, WAV, M4A/AAC, OGG/Opus

Note: your browser resampled this file, so the very top of the frequency scale may be approximate.

Check a whole folder

Scan an entire music folder at once — files are read locally, nothing is uploaded. On Chrome and Edge you can also move or delete the flagged files afterwards.

How Transcode Detection Works

Lossy codecs like MP3 throw away high frequencies. When someone re-encodes an MP3 into FLAC, the FLAC is technically lossless but only contains what the MP3 had — a fake lossless file. On a spectrogram this shows as a hard, flat lowpass cutoff with empty black space above it, instead of the natural energy that a real CD rip carries up to about 22 kHz.

Rough cutoff fingerprints: MP3 320 kbps ≈ 20.5 kHz, V0 ≈ 19.5 kHz, V2 ≈ 18.5 kHz, 192 kbps ≈ 19 kHz, 128 kbps ≈ 16 kHz. Genuine CD-sourced lossless reaches ~20–22 kHz with a soft rolloff.

Fake Hi-Res

A real 24-bit / 96 kHz file should contain audio above 22 kHz. If a "hi-res" file hits a wall at 22 kHz (or lower), it was upsampled from a CD — or worse, from a lossy source. This tool flags that too.

Beating "Audio Optimizers"

Some tools (like Platinum Notes) paste faint high-frequency noise above the real cutoff to fool checkers that just look for "any energy up high". This analyzer instead finds the sharp energy step and measures the flat, non-musical texture of the added noise — so it still spots the true cutoff hidden underneath.

Fake 24-bit

A file can claim 24-bit while its samples are really only 16-bit, padded with zeros. The analyzer checks whether the sample values are quantized to 16-bit steps and flags padded "fake 24-bit" files.

Checking a Whole Folder

Use the folder mode to scan an entire library at once and get a sortable table of results. Export a CSV report or an .m3u playlist of just the flagged files. On Chrome and Edge you can also move the flagged files into a /_suspect subfolder (reversible) or delete them — everything stays on your device.

Supported Formats

MP3, FLAC, WAV, M4A/AAC, and OGG/Opus decode in all modern browsers. The file is decoded at its native sample rate so the frequency ceiling stays accurate, and analysis happens entirely on your device.

How to Read a Spectrogram

Experienced listeners judge by eye rather than trusting any automatic verdict — automated detectors (including the well-known desktop ones) are notorious for flagging genuine tracks and missing real fakes. Here's what to look for:

  • Sharp brick-wall cutoff — a hard, flat horizontal edge with empty black above it is the classic lossy fingerprint. A genuine recording fades out gradually.
  • MP3 "blocks" — the definitive tell. Zoom into a ~1 second window: lossy encoding leaves a faint blocky, gridded texture in the high frequencies. You can only see it at high zoom, which is why this tool lets you drag-select a region to inspect it up close.
  • Horizontal lines are NOT lossy artifacts — they're usually hum or electrical interference from analog equipment. Don't mistake a thin high line for genuine full-band content.
  • Noise is broadband — real high-frequency energy (or dither) spreads across a range of frequencies, never a single line.
  • Use a linear frequency scale (this tool does by default) and compare both channels — some fakes differ between left and right.

Important: This Is Evidence, Not Proof

Some genuine recordings are naturally band-limited — old analog masters, certain acoustic material, or particular mastering choices. A gradual low cutoff is reported as "inconclusive" for exactly this reason. The automated result is a screening hint; the spectrogram image is the real arbiter, so read it with your own eyes.

Frequently Asked Questions

Is my audio uploaded anywhere?
No. Decoding and analysis run entirely in your browser using the Web Audio API. The file never leaves your device.
Can it tell a 320 kbps MP3 from a real FLAC?
Usually yes — a 320 MP3 shows a cutoff near 20.5 kHz while a genuine CD rip carries energy to ~22 kHz. The margin is small, so the spectrogram image is the final word.
Do I need the original file to compare against?
No. The lossy fingerprint (the cutoff) is visible in the suspect file on its own.
What does "window size" mean?
It's the FFT size — how many audio samples are analysed at once for each column of the spectrogram. It's a trade-off between time and frequency detail: a smaller window (2048) pinpoints when things happen more precisely (better for spotting MP3 block artifacts), while a larger window (8192) resolves which frequencies are present more precisely (a sharper, cleaner cutoff line). 4096 is a balanced default. It only changes how the spectrogram is drawn, never the audio itself.
Why does it say "evidence, not proof"?
Automated detection has false positives: some legitimate recordings are genuinely band-limited. Always confirm by looking at the spectrogram.