Krisp, best known for silencing background noise on Zoom calls, has just launched a real-time AI Accent Conversion tool designed specifically for Latin America, marking the first time voice tech of this kind is being deployed at scale in the region.
Rather than tweaking pronunciation or training agents to speak with a “neutral” accent, Krisp’s tool dynamically modifies live speech to match regional English inflexions, all while keeping the agent’s natural tone intact. That means agents can stop mimicking foreign accents and still be clearly understood by customers from Mexico to Colombia.
The new solution covers five major English-speaking Latin American accents, representing around 85% of the Spanish-speaking population in the region. It spans dialect groups including Mexican, Caribbean, Central American, and Andean/Neutral Spanish, allowing contact centres to localise conversations without sacrificing authenticity.
Removing Accent Barriers
As AI continues to flood contact centre workflows with promises of speed and efficiency, Krisp is aiming at something more human: removing barriers to understanding without erasing identity. Agents no longer have to fake accents to sound “acceptable.” Instead, they can focus on what they’re saying, not how they sound saying it.
By making calls more intelligible for customers, the tool sidesteps the subtle prejudices that can creep into interactions, while boosting agent confidence and reducing the need for agents to repeat themselves.
From a hiring perspective, the change is even more consequential. With accent no longer a limiting factor, companies can widen their talent pool and prioritise fluency and problem-solving over phonetics.
Krisp’s latest update also enhances voice naturalness and reduces accent leakage in its existing Indian and Filipino English models. More accents, including South African and non-U.S. The company plans to add other English varieties later this year.
BPM company Firstsource has recently deployed Sanas‘ Real-Time Accent Translation technology to transform spoken accents without altering tone or meaning.