STFT (short-time Fourier transform)
A Fourier transform taken over short, overlapping windows, so you learn not just which frequencies are present but when they appear.
A plain FFT over a whole song tells you there is 440 Hz in there somewhere, not where. The STFT chops the signal into chunks of perhaps 1024 or 2048 samples, multiplies each by a window function, slides forward by a hop and repeats. The output is a grid of complex numbers, time by frequency. The trade-off cannot be negotiated away: long windows give sharp frequency and blurry time, short ones the opposite. Dennis Gabor spelled this out in 1946.
With the right window and overlap the transform can be inverted by overlap-add, which makes it a workbench rather than just a meter. The phase vocoder, noise reduction and the mel spectrograms fed to speech models all run on it. The spectrogram is merely its magnitude, coloured in.