AssemblyAI has launched a text-to-speech API for voice agents at $15 per million characters. It offers 4,000 voices, instant voice cloning, pronunciation control and emotion tags. Supported languages are English, Spanish, French, German, Italian and Portuguese, not Korean. The company reports 90 ms to first audio in production and a separate 47 ms laboratory median; neither represents full speech-recognition-to-response latency. Pay-as-you-go access without an annual contract and concurrency terms are deployment conditions to verify. More advanced voice cloning is a future offering rather than part of the current launch.