Real-time AI voice technology alters accents in Indian call centers for better clarity

Teleperformance SE, the world’s largest call center operator, is implementing an artificial intelligence system designed to soften the accents of English-speaking Indian workers in real-time. The company claims this technology will enhance customer understanding and satisfaction.

“When you have an Indian agent on the line, sometimes it’s hard to hear, to understand,” Deputy Chief Executive Officer Thomas Mackenbrock told Bloomberg. “[This technology can] neutralize the accent of the Indian speaker with zero latency, [creates] more intimacy, increases the customer satisfaction, and reduces the average handling time.”

The technology analyzes speech input and modifies it to match a specified accent while preserving the speaker’s original voice and emotion. The system uses speech recognition to capture the speaker’s voice in real-time. Advanced algorithms then transcribe the spoken words into text, accounting for various accents and speech patterns. Once the system transcribes the speech, the core accent translation process analyzes the speaker’s accent and pronunciation patterns. Then, it applies AI models trained on datasets of various accents to modify the voice.

During the conversion, the software modifies intonation, stress patterns, and phoneme pronunciation to align with the target accent. The system employs text-to-speech (TTS) technology to synthesize the phonetic pattern and convert the text into a synthesized voice that maintains the speaker’s original tone, emotion, and identity.

Teleperformance is deploying technology with background noise cancellation in Indian call centers, providing customer support for international clients. Its customers include major tech companies like Apple, TikTok, and Samsung Electronics.

The technology is part of a larger Teleperformance AI investment strategy. The company plans to invest up to €100 million ($104 million) in AI partnerships in 2025. Palo Alto-based startup Sanas developed the accent technology after a $13 million Teleperformance investment earlier this year.

The software comes as the call center industry faces challenges from the rise of AI chatbots. Last year, Teleperformance experienced a significant drop in share prices after Swedish fintech firm Klarna Bank announced that its AI assistant was doing the work equivalent to 700 full-time agents. In response, Teleperformance has focused on using AI to enhance rather than replace its workforce of 490,000 employees.

Sanas states that its goal is to reduce “accent-based discrimination.” However, while the software may not replace workers directly, there are concerns about the potential impact on call centers in places like the Philippines, which have built their market position on high-quality English speakers.

Currently, the software supports Indian and Filipino inflections. Sanas is developing versions for other regions, including Latin America. Mackenbrock emphasized that while AI will be ubiquitous, “the human element will be incredibly important” in building connections and enhancing customer experiences.