DeepL has so far mostly been about translating written texts, but in an increasingly competitive and inventive AI market, the company is now launching live translations to help people host video chats in multiple languages. — Photo: Florian Schuh/dpa
COLOGNE: The German AI start-up DeepL is launching a feature for translating spoken language in real time, DeepL Voice, the company's chief executive and founder, Jaroslaw Kutylowski, has announced.
DeepL will then be able to provide live translations for video chat, allowing people to join the likes of a Zoom meeting in several languages. There will be an app for mobile devices that can be used to translate face-to-face conversations in real time.
DeepL is already a leader in the translation of written content, Kutylowski said at a company event in Berlin on Wednesday. However, the live translation of spoken content is a completely different matter.
"Sentences are sometimes not completed, pronunciation is often unclear and the latency cannot be too high." These limitations could lead to mistranslations and a poor user experience.
"These are the same challenges that can lead to misunderstandings in face-to-face interactions," said Kutylowski.
DeepL says its live translation feature is specifically designed for spoken language and can ensure smooth communication in multiple languages and to help overcome language barriers.
The company has been holding its own against strong competition, notably Google, which has a larger AI tool of its own, Gemini, which also supports real-time translations. Microsoft also offers live translation for its Teams video conferencing application.
However, the Cologne-based company has managed to compete with Google Translate and other offerings from major US providers in the past thanks to its relatively high translation quality.
Germany's most valuable AI start-up recently expanded its product portfolio to include a writing assistant that relies on artificial intelligence (DeepL Write Pro).
In July, DeepL launched a new generation of its language model with what it said was an increase in quality of machine translations. The company pointed to blind tests with language experts, showing that DeepL translations based on the new Large Language Model (LLM) were rated better than the results of systems from US tech giants. – dpa
