Rime

Conversational speech models

Rime develops voice models for conversational applications. Its product catalog distinguishes text-to-speech models from Prism, a speech-to-speech system that exposes intended reply text before producing audio. [1]

Rime launched Prism on October 28, 2026, describing an inspectable checkpoint for application rules and continued listening during spoken replies. The launch offered a trial while stating that self-serve reinforcement-learning and supervised-fine-tuning APIs were being built with beta customers. [2]