What is this?
AI audio spans three revolutions at once: music generation from a text prompt, voices indistinguishable from humans, and real-time conversation. It has produced chart-topping AI songs, an explosion of audiobooks and voice agents — and the first major copyright verdicts.
Key tools & players
- Suno, Udio — full songs (vocals, lyrics, production) from a prompt
- ElevenLabs — the leader in lifelike speech and voice cloning
- OpenAI / Google voice modes — real-time conversational voice
- Open source: MusicGen (Meta), Bark, Whisper for transcription
Milestones
- 2022 — Whisper makes near-perfect transcription free
- 2023 — MusicGen; viral "fake Drake" song forces the industry to react
- 2024 — Suno and Udio ship radio-quality songs; major labels sue both
- 2025 — Real-time voice agents replace call-center queues; voice cloning scams rise
- 2026 — Munich court rules against Suno in Europe's first big AI music copyright case (our coverage)
- Jul 2026 — OpenAI embeds Google DeepMind's invisible SynthID watermark in ChatGPT voice audio and opens a verification API — rivals converge on one provenance standard (our coverage)
- Aug 2026 — ByteDance's SeedRealtime takes full-duplex audio-visual: it watches, listens and speaks at once, live in the Doubao app
- Aug 2026 — Spotify badges "AI Persona" acts and drops them from recommendations by default: the first big platform to make being an AI artist cost something (our coverage)
Mini glossary
- TTS: text-to-speech — turning writing into natural voice
- Voice cloning: reproducing a specific person's voice from short samples
- Stem: an isolated track (vocals, drums) within a song