Scaled Cognition Inc. has launched Voice, a text-to-speech model designed for enterprise customer service. The tool supports over 25 languages and includes voice cloning capabilities. It is compatible with existing AI agents, contact centers, and legacy telephone systems.
The model features Smart Pronunciation, which interprets sentence context to accurately read addresses, confirmation codes, and dates. It can operate independently or integrate with the company's APT model to resolve ambiguities in customer conversations.
Pricing for the hosted API is $9 per million characters. For self-hosted deployments, the cost is $1 per million characters, excluding infrastructure expenses. The company estimates total costs at $13 per million characters for 500 million characters, dropping to under $7 at 1.5 billion characters.
Scaled Cognition reports a speech-generation latency of 70 milliseconds. The company previously raised $100 million in a Series A funding round led by Khosla Ventures LLC.
