Digital Audio 101 for AI Engineers

Audio systems turn changing air pressure into sampled numbers, transform those numbers into model inputs, and eventually produce a waveform for playback. Each representation carries assumptions about time, amplitude, channels, and frequency. Those assumptions affect both the sound and the system's storage and compute requirements.

Sampling rate: measurements over time

Sampling rate is the number of samples per second, expressed in hertz. A mono recording at 16 kHz contains 16,000 amplitude samples each second. Samples are discrete measurements of the signal at regular intervals.