Global artificial intelligence (AI) translation solutions company DeepL expanded beyond text translation into real-time voice translation, moving to target the enterprise language AI market. Its strategy is to establish itself not just as a simple translation tool but as "language infrastructure" applied across meetings, worksites, and customer service.
On the 15th at Co: society Seongsu in Seongdong District, Seoul, DeepL held "DeepL Connect Seoul" and unveiled its real-time voice-to-voice translation technology "Voice-to-Voice." This was the first time DeepL held this event in Korea, and the first case of running an entire event in multiple languages based on voice translation technology.
Gonçalo Gaiolas, DeepL chief product officer (CPO), said, "In a global corporate environment, language is no longer just a communication issue; it is a factor that affects overall organizational operations," adding, "Translation is shifting from a support function to core infrastructure for corporations to operate in real time."
The core technology DeepL unveiled is real-time speech-to-speech translation. Going beyond existing text translation or subtitle-based functions, a user speaks in their own language and the other party immediately hears it in their own language.
The products unveiled this time include: ▲ "Voice for Meetings," which supports real-time translation on video conferencing platforms ▲ "Voice for Conversation," for in-person conversations on mobile and web ▲ "Group Conversation," for multi-party translation in education and on-site settings ▲ "Voice-to-Voice API," applied directly to corporate systems.
With this, DeepL plans to extend its scope of application beyond meetings to factory floors, education, and customer centers. In the actual presentation, a scenario was presented in which voice translation is applied across the entire process—from meetings to production-site training to customer consultations—as a global company prepares for a product launch.
CPO Gaiolas explained, "Language issues affect overall corporate operations in various forms, such as meeting delays, slower decision-making, and delays in customer response," adding, "When small delays repeat, they have a significant impact on execution speed and quality."
According to DeepL's own survey, 90% of workers at Korean corporations said voice translation is necessary for their work, but the actual usage rate was only 36%. In addition, 70% said they have difficulty conveying meaning during translation, and 64% said they hesitate to express their opinions at all.
DeepL sees closing this "demand–utilization gap" as the key. The company aims to enable members to participate without language burdens through real-time voice translation. CPO Gaiolas said, "The moment people speak less or simplify because of language, the quality of a corporation's decision-making declines," adding, "True expertise emerges only when people can communicate naturally in their native language."
On technological competitiveness, the company emphasized differentiation from general-purpose Generative AI. Although general-purpose AI such as ChatGPT and Gemini has been rapidly advancing its translation features, it still has limitations in corporate environments, the company said.
He said, "General-purpose AI is optimized for a variety of uses, but in corporate environments, accuracy, consistency, brand tone maintenance, and system integration are important," adding, "DeepL is years ahead as a platform specialized for mission-critical translation."
DeepL also said it operates a hybrid architecture through its own GPU (graphics processing unit) cluster-based infrastructure and collaboration with AWS (Amazon Web Services), and that it trains its models using thousands of language experts. Corporate customers can build customized translation models that reflect glossaries and style guides.
The company is also improving issues currently cited as technical limitations, such as latency and naturalness of speech. DeepL plans to introduce a "voice cloning" feature this year that preserves the speaker's voice. It is also expanding investment in the Korean market. CPO Gaiolas said, "Korea is a strategic market with a fast pace of AI adoption," adding, "We are also reviewing ways to place infrastructure closer to the local market in line with future demand growth."
Starting with this voice translation technology, DeepL plans to expand into an "end-to-end language AI platform" connecting text translation, voice, documents, and content management. CPO Gaiolas said, "Now corporate competitiveness comes from expertise, not language ability," adding, "If you remove language barriers, organizations can move faster and make better decisions."
Meanwhile, on the real-time interpretation market, CPO Gaiolas said it will "expand the market rather than replace human interpreting." He said, "Professional interpreting will continue to be needed in high-risk, high-precision areas, but in most work environments, usage has been limited by expense and scalability," adding, "AI interpreting will cover the 90% of areas that have not been addressed so far."