krisp

Real-time, enrollment-free, and dramatically more natural.

 

We’re excited to announce the launch of v3 of Accent Conversion technology, a major technical advancement in Krisp’s voice AI stack. This version delivers significant improvements in voice naturalness, phoneme-level precision, and speaker adaptation—all while maintaining our commitment to security, scalability, and ease of use.

A leap in performance, now patented

With v3, Krisp’s Accent Conversion has reached a new level of technical sophistication—and is now officially patented. This milestone validates our unique approach to accent transformation: one that preserves speaker identity, operates in real time, and avoids the privacy pitfalls of voice cloning or enrollment-based models.

Accent barriers have long created friction in global contact centers, leading to miscommunication, increased call handle times, and hiring limitations. Krisp’s patented technology instantly solves these challenges, making conversations more natural and bias-free.

 

“With v3, we’ve dramatically improved one of the toughest challenges in voice AI—adjusting an accent while keeping the speakers voice and natural tone intact,” said Davit Baghdasaryan, CEO and Co-Founder at Krisp. “Now backed by patents, it brings us closer to voice technology that feels human and helps global support teams build stronger connections with customers.”

Hear the difference: v2 vs v3

To make the technical improvements tangible, we’ve created recordings using Indian and Filipino English speakers across the current and previous versions of Accent Conversion.

Agent Original speech AC V2 AC V3
Mandovi (IN accent)
Sherwin (IN accent)
Princess (PH accent)
Louis (PH accent)

The leap in quality is clear. Where v2 introduced early gains in clarity, v3 offers a step-change in naturalness—with smoother tone and intonation, fewer sound artifacts, and more accurate voice preservation.

Real-time voice adaptation—no setup required

One of the notable enhancements in v3 is its zero-shot deployment and usage. The technology doesn’t require:

 

  • Voice enrollment by the agents
  • User training
  • Configuration or tuning

 

Simply connect your headset, enable Accent Conversion, and start speaking. Voice adaptation is instantaneous and speaker-agnostic—meaning if another person picks up the headset, the system automatically recalibrates in real time, with no drop in quality or smoothness of the flow.

 

This level of usability sets a new standard for production-ready, plug-and-play technology.

Privacy by design: no voice data stored

Accent Conversion v3 is built with privacy and security at its core.

 

  • No voice embeddings are stored
  • No personal voice data is saved on-device or in the cloud
  • All processing happens on the spot, in real time

 

This design not only simplifies compliance with all the stringent security protocols—it also eliminates security concerns related to voiceprints or long-term data storage.

Phoneme-level precision and naturalness

From a technical standpoint, v3 introduces marked improvements across core speech processing components, including:

 

  • Finer phoneme-to-phoneme mapping
  • Improved prosodic modeling
  • Substantially reduced roboticness and speech artifacts
  • Greater emotional and inflectional nuance retention

 

As a result, agents sound more human, more authentic, and much easier to understand—especially in high-stakes communication environments like customer support, where comprehension and clarity are non-negotiable.

 

Compared to v2, v3 demonstrates:

 

  • Significant gains in intelligibility
  • Sharper articulation of consonants and vowels
  • A more fluid and natural speech cadence

A global roadmap ahead

Accent Conversion v3 currently supports a growing list of accents, with new geographies on the horizon. Our roadmap includes:

 

  • Latin American English accent pack
  • South African English accent pack
  • Conversion to regional U.S. English accents

 

Our goal is to support truly global teams with technology that adapts to real-world linguistic diversity—while retaining technical performance and security across all environments.

 

Accent Conversion v3 is now available for integration into your Krisp-powered workflows.

Check out more here: https://krisp.ai/accent-conversion/

Related Articles