Krisp's new AI flips Indian accents to American on live calls

Audio startup Krisp just unveiled a real-time accent conversion feature aimed squarely at global remote workers. Starting with Indian English, the software automatically translates spoken accents into standard American English mid-call.

Krisp says the technology keeps the speaker’s unique voice timbre intact while swapping out specific phonemes to match American pronunciation patterns. The feature is now rolling out in beta on Krisp's desktop app, where users can toggle it on or off before or during a meeting.

Real-time accents without the training

According to co-founder Arto Minasyan, the tool doesn’t require any voice training. It builds an acoustic profile of the speaker on the fly, relying on thousands of diverse speech samples alongside consent-backed meeting data Krisp gathered from its assistant tool.

Minasyan said the project started two years ago out of personal frustration when people struggled to understand his own accent despite fluent English. Krisp targeted Indian English first because Indian professionals represent an outsized share of the global STEM workforce, with Filipino English set to arrive next.

On paper, the business pitch looks lucrative. Krisp claims enterprise pilots saw sales conversion rates climb 26.1%, with revenue per closed book jumping 14.8%.

Yet the technology clearly has rough edges. A hands-on test by TechCrunch revealed that processed speech sounded unnatural and occasionally dropped words altogether. Krisp blamed early beta growing pains and promised ongoing model updates.

Krisp, which last raised venture capital in 2021, plans to roll out mobile apps on iOS and Android later this year for in-person meetings, alongside a dedicated Chrome extension for Google Meet.

Why it matters

For international sales reps and remote engineers, accent bias remains an unspoken hurdle in closing deals and managing clients. If Krisp can reliably iron out phoneme mismatches, it could meaningfully cut down on meeting friction. But based on early feedback showing robotic cadence and missed words, relying on it for high-stakes business calls today carries serious risk. Dropping a key technical spec or pricing figure is far worse than speaking with an authentic regional accent.

On the competitive front, AI voice tech is shifting from passive background noise removal toward real-time voice identity alteration. Rivals like GV-backed Sanas have already targeted offshore call centers at scale, but Krisp is bringing this directly into the mainstream remote worker desktop.

We think Krisp’s zero-setup approach is technically clever because eliminating user voice onboarding makes adoption seamless. However, the elephant in the room isn't just latency and audio artifacts; it’s the optics. Tech that actively scrubs foreign accents to appease Western ears will always trigger uneasy debates over workplace assimilation versus actual inclusion.