Introducing AssemblyAI-TTS-1: text-to-speech built for voice agents
AssemblyAI launches AssemblyAI-TTS-1, its first text-to-speech model for voice agents, streaming audio in about 90 ms with unlimited concurrency and no annual contract, at $15 per million characters. It offers 4,000+ voices, instant cloning and six launch languages.
What the article says
AssemblyAI, a voice AI company, has launched AssemblyAI-TTS-1, its first text-to-speech model, described on October 27, 2026 as generally available. It targets voice agents on live phone calls, where slow or unnatural speech makes callers hang up. The company reports audio streaming in 90 ms in production (47 ms lab median), a price of $15 per million characters, unlimited concurrency on pay-as-you-go, more than 4,000 voices, instant voice cloning, and accents in six languages.
The page compares median latency (47 ms versus 119 ms for ElevenLabs Flash v2.5 and 143 ms for Cartesia Sonic 3.6) and reports 93.7 percent pronunciation accuracy on 580 hard targets such as phone numbers. These are the company's own figures, supplemented by measurements from Coval, Cekura and Artificial Analysis. Limits: no fine-tuned cloning yet, no SSML or speed parameter; pronunciation dictionaries and emotion tags are listed as coming soon. The company recommends testing the model on a slice of live traffic before switching over.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of AssemblyAI. The content and rights belong to the publisher: please read the original article.