Scaled Cognition Launches Voice Model for Customer Service

Scaled Cognition introduces Voice, a speech synthesis model for customer service, according to SiliconANGLE. The model supports more than 25 languages, voice cloning and integration with AI agents, contact center platforms and older telephone systems.

Voice is available through a hosted API for $9 per million characters or as a self hosted model for $1 per million characters, excluding infrastructure costs. Scaled Cognition says one GPU can support about 32 simultaneous conversations and that speech generation begins in about 70 milliseconds.

A Smart Pronunciation feature interprets context before producing audio, helping the model read abbreviations, addresses, account numbers and dates. Voice can also work with Scaled Cognition's APT customer service model, which supplies conversation context to resolve ambiguous text, but APT is not required.