Krisp, an audio startup, has launched a new AI-powered feature designed to modify a user’s accent during calls. Initially available in beta for its desktop app, the tool specifically converts Indian English to US English while preserving the speaker’s natural voice by only adjusting the phonemes.
The idea emerged when co-founder Arto Minasyan, frustrated by frequent misunderstandings due to his accent, decided to develop a solution that could improve communication. After two years of development, the beta version is now accessible to users who can activate the feature at any point during a call.
Although initial tests revealed that the processed voice sometimes sounded unnatural and occasionally skipped words, Krisp attributes these issues to the early release and assures improvements in future updates. Enterprise trials have shown promising results, with a 26.1% increase in sales conversion rates and a 14.8% boost in revenue per booking when the feature was used.
Targeting Indian accents first was a strategic choice, as a significant portion of the global STEM workforce hails from India. The company also plans to expand support to other accents, including Filipino, as it refines the technology. Krisp trained its model on thousands of speech samples representing a variety of accents and dialects, utilizing data from its meeting assistant with proper user consent.
One notable advantage of the tool is its ability to create a real-time profile of the speaker’s voice without any need for pre-training. In addition to the desktop app, Krisp intends to roll out iOS and Android applications later this year to support in-person meetings, along with a new Chrome extension aimed at enhancing integration with Google Meet.
This innovation is part of a broader trend in AI-driven voice modification, aiming to improve communication by making accents more universally understandable while preserving the uniqueness of each individual’s voice.
