Audio Spectral Forensics: How to Detect Fake Lossless & MP3 Transcodes
In the digital music distribution era, the demand for high-resolution lossless audio (FLAC, ALAC, WAV, and AIFF) has never been higher. Unfortunately, digital audio piracy, grey-market music vendors, and automated re-upload bots have created a widespread problem: the fake lossless transcode (often referred to in audiophile and DJ communities as an up-converted or upscaled track).
A transcode occurs when someone takes a low-bitrate lossy file (such as a 128 kbps MP3 ripped from YouTube, a 96 kbps voice note, or an old peer-to-peer download) and re-encodes it into a lossless .flac or 16-bit .wav file. While operating systems and media players will proudly report "1,411 kbps 16-bit FLAC", the underlying audio data remains permanently degraded. Lossy psychoacoustic compression is mathematically irreversible: once acoustic frequencies and fine transient details are discarded by a lossy encoder, no software can ever regenerate them.
1. The Physics of Lossy Audio Compression: Why Cutoff Shelves Exist
Lossy encoders like MP3 (MPEG-1 Audio Layer III), AAC (Advanced Audio Coding), and Ogg Vorbis rely on sophisticated psychoacoustic masking models. Because human hearing is less capable of perceiving quiet sounds in the presence of loud adjacent frequencies, the encoder discards inaudible data to drastically reduce file sizes.
However, when bitrates are restricted (e.g., 128 kbps), the encoder does not have enough bits to represent the entire audible spectrum ($20 ext{ Hz}$ to $20 ext{ kHz}$). High frequencies consume an enormous proportion of the data budget because rapid air vibrations require numerous sample transitions. To conserve bits for the critical human speech and midrange frequencies ($200 ext{ Hz}$ to $4 ext{ kHz}$), lossy encoders apply a steep brickwall low-pass filter, completely severing all sound above a fixed cutoff frequency.
2. The Exact Frequency Cutoff Signatures of Common Audio Formats
By inspecting an audio file with a high-resolution Fast Fourier Transform (FFT) waterfall spectrogram, you can instantly determine the file's true origin based on its high-frequency cutoff shelf:
| Audio Format / Bitrate | Standard Cutoff Frequency | Visual Spectrogram Characteristics | Authenticity Rating |
|---|---|---|---|
| 128 kbps MP3 (Standard Web / YouTube) | 16.0 kHz |
Sharp, razor-flat brickwall cutoff at 16 kHz. Total pitch-black silence above 16 kHz. | LOW QUALITY |
| 192 kbps MP3 (Medium Quality) | 18.5 kHz — 19.0 kHz |
Cutoff shelf between 18.5 kHz and 19 kHz. Occasional transient spikes may punch through. | MID TIER |
| 320 kbps MP3 (LAME CBR/VBR) | 20.0 kHz — 20.5 kHz |
Gentle low-pass roll-off starting at 20 kHz. Minimal activity between 20.5 kHz and 22.05 kHz. | AUTHENTIC MP3 |
| 256 kbps AAC (Apple Music / iTunes) | 20.0 kHz — 21.0 kHz |
Sophisticated MDCT filter bank. Preserves natural harmonics better than MP3 at identical bitrates. | HIGH-GRADE LOSSY |
| True Lossless (16-bit 44.1 kHz FLAC/WAV) | 22.05 kHz (Nyquist Limit) |
Continuous harmonic overtones, cymbal sizzle, and transient room acoustics extending all the way to 22.05 kHz. | GENUINE LOSSLESS |
| High-Res Studio Master (24-bit 96 kHz) | 48.0 kHz (Nyquist Limit) |
Ultrasonic harmonic overtones extending beyond human hearing (>20 kHz) up to 48 kHz. | STUDIO MASTER |
3. What is Stereo Phase Correlation and Why Does it Matter?
A great master does not simply sound fantastic in stereo headphones; it must also translate flawlessly onto single-speaker playback systems. When two stereo channels (Left and Right) are summed to mono, any acoustic signals that are $180^\circ$ out of phase will cancel each other out through destructive interference ($ ext{Left} + ext{Right} = 0$).
The Stereo Phase Correlation Meter evaluates the mathematical relationship between the Left and Right channels on a scale from $-1.0$ to $+1.0$:
- $+1.0$ (Pure Mono): The Left and Right channels are identical in both phase and amplitude. Flawless mono translation with zero phase cancellation.
- $+0.5$ to $+0.8$ (Commercial Stereo Sweet Spot): Natural acoustic stereo imaging with rich spatial width and robust mono compatibility.
- $0.0$ (Complete Stereo Independence): Left and Right contain completely uncorrelated acoustic material (e.g., dual mono tracking).
- $-1.0$ to $0.0$ (Out-of-Phase / Cancellation Warning): Severe anti-phase information. When played back on mono club sound systems, smartphone speakers, or smart home assistants, bass frequencies and lead vocals will vanish into thin air.
4. Frequently Asked Questions (FAQ)
Can a genuine, authentic FLAC file ever have a 16 kHz cutoff?
Yes, in rare specific historical scenarios. If you analyze a song recorded in the 1960s or 1970s on vintage magnetic analog tape, or an early 1980s digital synthesizer with a 32 kHz internal clock (such as early Fairlight CMI or E-mu samplers), the original master recording never contained frequencies above 15 kHz or 16 kHz.
However, for modern commercial music produced after 1995, an abrupt horizontal shelf at 16 kHz is nearly 100% indicative of an illegal or deceptive transcode from a 128 kbps MP3 source.
Why does 320 kbps MP3 have a cutoff at 20.5 kHz instead of 22.05 kHz?
The LAME MP3 encoder (the gold standard for MP3 encoding) intentionally applies a low-pass filter at approximately 20.5 kHz even at maximum 320 kbps CBR. This is by design: human hearing barely perceives frequencies above 20 kHz, so discarding that tiny sliver of ultrasonic air allows the encoder to allocate more bits to the critical audible spectrum between 1 kHz and 16 kHz, resulting in fewer audible artifacts.
How does this online tool differ from desktop software like Spek or Audacity?
Desktop tools like Spek require downloading, installing, and managing native executable binaries. The Studio LOOS Analyzer runs 100% in your web browser with zero installation, zero server uploads, and zero cross-platform compatibility issues. It leverages a custom Cooley-Tukey Radix-2 FFT engine implemented in JavaScript typed arrays to compute the entire track's spectrogram in milliseconds.