Audio is just another sequence — a 1D signal sampled at 16 000 or 44 100 times per second. Modern audio models convert raw waveforms into a time-frequency representation called a spectrogram, then apply the same transformer architectures that power LLMs and image models. This lesson covers the three dominant audio generation paradigms: automatic speech recognition (Whisper), text-to-speech (TTS), and music generation (MusicGen).