Voice AI company Krisp has unveiled VIVA 2.0, an upgraded version of its voice infrastructure platform aimed at making AI-powered voice agents more reliable in real-world conditions. The latest update brings a new set of lightweight, real-time audio models designed to help conversational AI systems better handle interruptions, recognize accents, detect synthetic voices, and determine when a user has finished speaking.
The launch comes as businesses rapidly expand the use of voice AI for customer support, IVR systems, and automated service operations. Krisp noted that voice agent adoption surged ninefold in 2025, though many AI voice systems still face challenges in noisy or unpredictable environments where speech interruptions and recognition errors are common.
Robert Schoenfield, EVP of Licensing and Partnerships at Krisp, said voice is quickly becoming the main interface between humans and AI, but real conversations rarely happen in perfectly quiet settings. He emphasized that AI systems must be able to adapt to natural human interactions and background noise.
Krisp explained that VIVA 2.0 functions as an infrastructure layer between speech-to-text engines and large language models, helping clean and interpret audio before it reaches the AI system.
Among the major upgrades is Turn Prediction v3, a multilingual model that can identify when a speaker has completed a sentence without depending on text transcription. The company said this helps conversations feel smoother by reducing awkward interruptions.
The platform also introduces Interrupt Prediction, which can tell the difference between actual interruptions and simple acknowledgements like “hmm” or “yes,” allowing voice agents to respond more naturally.
In addition, Krisp added new real-time signal detectors capable of identifying accents, speaker gender, and AI-generated speech. The company said the accent detection system can direct audio to speech recognition models optimized for specific accents, improving transcription accuracy.
Another key enhancement is Voice Isolation v3, which is designed to cut background noise and improve speech clarity during calls.
According to Krisp, the platform now handles more than 12 billion minutes of voice AI traffic each year and is already integrated into over 130 voice AI products, including platforms such as Telnyx, LiveKit, and Vapi.
Also Read: Shunya Labs Unveils Sovereign AI Platform for Enterprise Voice Systems
