DeepL launches DeepL Voice, real-time voice translator

DeepL is a German startup that ended up becoming well known thanks to its accurate and detailed translations in its products. Therefore, the company has plans to stand out even more in this area, and now it brings something very interesting with the launch of DeepL Voice.

DeepL, which currently has a $2 billion valuation and more than 100,000 paying customers, is now betting on real-time audio translation with the new feature. The idea is to bring a more accurate and faster option to the public in this segment. Check out more details.

Also read:

Galaxy Buds headphones can now translate calls in real time
Telegram gains real-time message translation, emojis with full-screen animations and more

DeepL Voice translates voice to text in real time

The new DeepL Voice feature allows users to hear someone speaking in another language and automatically receive the translation in their own language (or another chosen language), all in real time and with the maximum accuracy possible.

DeepL is currently capable of capturing audio in languages ​​such as English, German, Japanese, Korean, Swedish, Dutch, French, Turkish, Polish, Portuguese, Russian, Spanish and Italian, offering translated subtitles in up to 33 languages, which shows a large scope.

The feature is primarily aimed at live meetings and video conferencing, but goes beyond common audio translation. This is because, instead of generating the result in audio or video, DeepL Voice presents the translations in text form. This ensures faster processing, which is important for live conversations, as many AI-based translation solutions still experience some delay, which ends up affecting real-time interactions.

How DeepL Voice works

In the case of face-to-face meetings, the user can configure the resource on a smartphone, placing it between participants at a meeting table so that both see the translation as a “mirror image”.

In videoconferences, translations appear in the form of subtitles, a feature that is currently only available on the Microsoft Teams platform, which already reaches the majority of DeepL’s B2B clients. Jarek Kutylowski, founder and CEO of DeepL, sees the new feature as the company’s first step into voice translation and says this is the future of translation for the next year.

Although the main focus is for use in corporate meetings, DeepL still believes that the resource can also be put to good use in other areas, such as customer service, in places such as restaurants and other establishments that serve directly the public. This is because it would ensure more efficient communication between employees and customers who speak different languages, contributing to better service and user experience.

DeepL’s difference lies in its technology

But as cool as the new feature is, it’s not exactly new, which means competition in the real-time translation market is at an all-time high. Companies like Google and startups specializing in AI are also exploring automatic audio translations, with tools like Google Meet’s automatic captions and Eleven Labs Dubbing, which uses deepfake voices to deliver translations synchronized with video.

Although voice AI solutions offer new possibilities, DeepL Voice’s difference lies in the speed of text processing. For Kutylowski, the text translation feature is calculated and displayed quickly, while audio and video processing technologies still need to evolve to deliver results as quickly as text.

Since 2017, when the company launched its translator, the demand for voice translation has been the main request from DeepL’s customers. Unlike many companies that use language models from other companies, DeepL develops its own translation technologies. In July, the company launched a new language model optimized specifically for translations, which Kutylowski says even outperforms OpenAI’s GPT-4 and Google and Microsoft’s translation solutions.

Privacy and data security

With the growth of AI technology, concerns about privacy and data protection are also growing. DeepL Voice transmits the audio to the company’s servers, where the translation is processed, but Kutylowski guarantees that no audio data is stored or used to train the company’s language models.

DeepL ensures that it is committed to following data protection regulations, including GDPR, and will work in partnership with its customers to prevent any breaches.

Fonte: techcrunch