Rime, a San Francisco voice technology company, launched Prism, a speech-to-speech model for enterprises that can be fine-tuned with customers’ data. The model generates audio directly while producing a verifiable text response before it is delivered. Prism responds to the first audio input in roughly 500 to 570 milliseconds and scored 56.6 on the tau-Voice benchmark, according to Rime. In a trial with Revmo, it completed orders with the caller’s consent 94% of the time, compared with 57% for an OpenAI model, while achieving a 100% order confirmation rate and 85% of calls free from hallucinations. The company, which powers nearly 100 million phone calls each month, targets healthcare, finance, airlines and food service, and announced a $35 million fundraising round, while previously disclosed rounds amounted to approximately $29.5 million to $32.6 million.
Source: Read the original article
