Speech & Audio AI — Roadmap

1. Audio Fundamentals

• Digital audio: sampling rate, bit depth, the Nyquist theorem
• The Fourier transform: time domain to frequency domain
• Spectrograms: the Short-Time Fourier Transform (STFT)
• Mel spectrograms and MFCCs (Mel-Frequency Cepstral Coefficients)

2. Speech & Audio Tasks

• ASR (Automatic Speech Recognition): CTC, encoder-decoder, and Whisper-style approaches
• TTS (Text-to-Speech): the acoustic-model + vocoder pipeline, end-to-end neural TTS
• Speaker recognition and verification
• Speaker diarization ("who spoke when")
• Audio classification (sound events, music genre)
• Speech enhancement and noise suppression
• Audio-language models and multimodal audio

Next: LLMs & GenAI — where audio-language models connect back to the general multimodal Transformer story.