Spectrogram · audio
Was this audio upsampled or re-encoded?
A spectrogram shows a recording frequency by frequency: time runs across, pitch runs up, and brightness is how much energy sits at each point. It is the fastest way to see a codec's tell, the sharp horizontal line where a lossy encoder or an upsample cut off the top of the sound.
Runs entirely in your browser. Your file is never uploaded or stored.
Decoding and analysing audio…
How to read it
Each vertical line is one slice of time; within it, low notes sit at the bottom and high frequencies climb to the top, up to half the sample rate. Bright means strong energy at that pitch and moment, dark means little or none. Speech and music fill the lower and middle bands with clear harmonic stripes; hiss, cymbals and sibilance live up high. A natural, full-quality recording fades gradually into noise at the very top rather than stopping dead.
On a linear scale almost everything you care about is squeezed into a thin band at the bottom, because a voice lives below a few kilohertz while the axis runs to twenty. The scale selector fixes that: logarithmic, mel, Bark and ERB all stretch the low and mid frequencies to fill the height, spreading the detail the way the ear hears it, while period plots time-per-cycle instead of frequency. Linear stays best for reading an exact cutoff height; the perceptual scales are best for seeing structure.
The frequency-cutoff tell
The most useful thing a spectrogram reveals is a brick-wall cutoff: a hard horizontal edge with near-black emptiness above it. Lossy encoders throw away the top of the spectrum to save space, so an MP3 shows a shelf around 16 kHz, and lower bitrates cut lower still. If a file claims to be a lossless or high-quality original but has a sharp shelf partway up, it was almost certainly made from a lossy source, or upsampled from a low sample rate to look like more than it is. A gentle roll-off is normal; a razor-straight line is the fingerprint of compression.
What it does not decide
A spectrogram is evidence to read, not a verdict. Some clean recordings genuinely have little high content, and some processing chains reshape the top end for innocent reasons. Treat a cutoff as a strong lead about a file's history and corroborate it with the metadata reader and the codec history before concluding.