Spectrograms: Sound as an Image

One Fourier transform tells you a sound's recipe — but not when each ingredient entered. So slice the waveform into overlapping chunks of ~25 milliseconds, FFT each chunk, and stack the results side by side. That's a spectrogram.

• X-axis — time
• Y-axis — frequency
• Color — energy at that frequency, at that moment

Here's the payoff: you can read it. Vowels paint bright horizontal bands; 's' and 'sh' smear energy high up; a dog bark is a short vertical burst. Speech becomes legible texture.