Artificial intelligence in voice translation overcomes the "uncanny valley" stage and reaches the level of natural communication with the preservation of emotions and intonations. Thanks to new technologies, people will be able to speak their native language freely without barriers using smart glasses or phones.
Artificial intelligence in voice translation overcomes the "uncanny valley" stage and reaches the level of natural communication with the preservation of emotions and intonations. Thanks to new technologies, people will be able to speak their native language freely without barriers using smart glasses or phones.
This was stated by Stefan Mesken, the chief scientific officer of DeepL, during the GITEX AI Europe 2026 conference in Berlin, Interfax-Ukraine reports .
According to Mesken, the turning point for AI-powered text translation occurred back in 2017–2018. For the past three years, developers have focused on applying this experience to spoken speech. This process has proven more difficult due to the need to convey not only the essence of the text, but also the emotional component and unique characteristics of the speaker's voice.
The scientist gave an example from a conference in Tokyo at the end of last year. At that time, the company tested its DeepL Voice product in parallel with live simultaneous translation.
“After about half an hour, professionals in the field realized that technology could already provide a better experience. Simultaneous interpretation by a human is a very difficult task and takes about 30 seconds. With technology, this is reduced to about 3 seconds. So you get a much better experience from live listening,” explained a company representative.
DeepL is currently focused on the next stage of technology development - translation with full preservation of emotional features. During the operation of the algorithm, the AI's voice initially sounds somewhat robotic, but within three seconds the system learns the features of human speech and adapts, copying its timbre and intonation. During a speech in Berlin, the developers demonstrated this function: English was translated into German online with an accurate reproduction of the voices of the participants on stage.
Mesken is confident that such a breakthrough will allow people and companies to work and do business more comfortably in their native language. Addressing one of the audience members in the hall, who said that he knows six languages, the scientist joked:
"Six languages? That's incredible. Give this man a round of applause. But it turns out it was a waste of time, my friend. Pretty soon you'll just be able to plug people into your glasses or something."
Earlier, the German startup introduced a new product — DeepL Agent , which is intended to automate routine processes in companies and become a competitor to solutions from OpenAI, Anthropic, and Microsoft.
