Rime Announces Prism, A Voice AI That Checks Its Own Output Before Speaking
Rime announced Prism, a speech-to-speech voice AI for enterprise use, on October 28, 2026. Prism understands a caller's voice directly and drafts its reply as text before speaking, so companies can check and correct the output against set rules.
1 source 35m ago
Reporting from 1 source: GIGAZINE.
Why it matters
Prism's design answers the tension Rime names in enterprise phone work: direct voice-to-voice models keep tone and hesitation but are hard to police, while text pipelines are controllable but lose those cues, and Prism tries to hold both by writing the reply before it speaks.
Key Facts
- Rime announced Prism, a speech-to-speech voice AI for enterprise use, on October 28, 2026.
- Prism outputs its response as text before speaking so companies can check and correct it.
- Rime says Prism responds to a user's voice in about 300 milliseconds.
- Prism scored a 56.6% task success rate in Rime's τ-Voice evaluation.
- In the telecom field test, Prism's task success rate was 82.5% against 45.6% for OpenAI's GPT-Live-1.
Rime says Prism responds in about 300 milliseconds and scored a 56.6% task success rate in its τ-Voice evaluation.
Prism keeps listening while it talks, so a caller can cut in mid-sentence and still be answered. Rime also built in tool calls, letting the AI pull customer information during the conversation, and customization for the pronunciation of names and account terms.
In Rime's τ-Voice test, built around airline, retail, and telecom inquiries, Prism's first voice output came at 0.57 seconds, about half the time of OpenAI's GPT-Live-1 and Google's Gemini 3.8 Live. In the telecom portion, Prism's task success rate was 82.5% against 45.6% for GPT-Live-1. Prism is on trial with some companies now; a general-purpose API is planned.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google