Scaled Cognition has introduced Voice, a text-to-speech (TTS) model designed for enterprise customer experience (CX) applications. The company says Voice delivers high-quality, natural-sounding speech with ultra-low latency at a lower cost than comparable solutions from leading providers. Scaled Cognition also developed APT, an AI model designed to improve reliability and reduce hallucinations in enterprise customer interactions.
Voice is available self-hosted at $1 per million characters, so enterprises no longer have to choose between the best-sounding voice and the one they can afford to run on every call.
Voice responds at human-level speed, with 70-millisecond speech-generation latency, fast enough that the voice is never the reason a caller hears dead air. Built-in text normalization gets the details right. For example, it reads “Pay 1/2 by 1/2” as “pay half by January second,” and automatically decides how to speak dates, addresses, account numbers and dollar amounts. Teams can fine-tune how it reads brand names and specialized terms.
Alongside the launch, Scaled Cognition is inviting buyers to compare Voice directly with leading TTS providers in a blind listening test. Participants hear five pairs of customer-service conversations and choose the voice they would prefer to hear without knowing which provider generated each clip.
“There are a lot of voice models out there, but none of them give you this quality and accuracy at this price,” said Dan Roth, CEO and co-founder of Scaled Cognition. “Enterprises have been told to pick the voice that sounds best or the one they can afford to run on every call. Now, with Voice, they don’t have to pick.”
Voice launches with more than 25+ languages and voice cloning, with additional languages planned. It works with existing agents and legacy systems, so enterprises can adopt it without moving to Scaled Cognition’s broader platform.
Paired with APT, Scaled Cognition’s model for CXs, Voice carries Super-Reliable Intelligence through to the words customers hear. APT handles the conversation and takes verified actions in enterprise systems within business policy, and Voice speaks the result, reading confirmation codes and account numbers exactly as APT verified them.
Voice is generally available today through the Scaled Cognition API at $9 per million characters, or self-hosted at $1 per million characters, well below what leading providers charge for comparable premium voices. Visit the voice product page and the blind listening test quiz to learn more.
Enterprises can get started and take the blind listening test at scaledcognition.com.
Related News:
Descope Agentic Identity Hub 3.0 Expands AI Agent Governance
