German company DeepL running an advanced translator she announced a few days ago a bunch of news. The first is the introduction of two new products for companies Voice for Meetings a Voice for Conversationfurther support voice input on computers a buying AI infrastructure from Nvidia.
Text dictation on computers
If you use DeepL on a mobile device, you are familiar with this feature. After all, it also makes more sense on mobile phones, given that you hardly have a computer with you during normal communication on the street. It is now possible to dictate the voice on computers (although Czech is still missing).
A pair of DeepL Voice products for business
A more significant innovation is the introduction of a pair of new products Voice for Meetings and Voice for Conversation. These will find their application especially in the corporate segment, where the company is also primarily aiming. Both features can translate spoken word into subtitles in real time using language artificial intelligence.
Voice for Meetings will translate conference calls and meetings into subtitles, Voice for Conversation will then be used for calls on smartphones. In this way, the company wants to remove language barriers in meetings and allow each participant to speak in the meeting or on the phone in their preferred language.
“DeepL is already a leader in text translation, but real-time speech translation is something else entirely. When translating ongoing spoken speech, you’re dealing with incomplete input, pronunciation complications, latency, and the like, which can lead to inaccurate translations and negatively impact user-friendliness. These are the same issues that we often encounter in normal conversation and that can lead to misunderstandings, so we created a solution that takes them into account from the start and allows businesses to overcome language barriers by allowing them to communicate in multiple languages as needed” says Jarek Kutylowski, CEO and founder of DeepL.
DeepL Voice real-time voice translation currently only supports English, German, Japanese, Korean, Swedish, Dutch, French, Turkish, Polish, Russian, Spanish and Italian, while subtitle translation is supported in all 33 DeepL languages (including Czech and Slovak). Support for additional languages for voice is to come in the future.
Where DeepL can’t, NVIDIA can help
At the end of October this year, the company was bought by Nvidia DGX SuperPOD system with DGX GB200 in order to train their models and improve features and products. The system should be available to DeepLu around the middle of 2025, and according to him, this is the biggest purchase the company has made during its existence. The DGX SuperPOD system contains 36 GB200 Grace Blackwell computers connected as one using Nvidia NVLink.
Source: www.cnews.cz
