DeepL, a renowned name in translation technology, has unveiled a groundbreaking voice-to-voice translation suite that caters to diverse scenarios, including meetings and group conversations. This innovative tool is designed for frontline workers and can be integrated into custom applications. Additionally, an API is being introduced, enabling developers and businesses to leverage DeepL's technology for tailored solutions, such as enhancing call center operations.
In an interview with TechCrunch, DeepL's CEO, Jarek Kutylowski, expressed that the transition from text to voice translation was a logical progression for the company. "We have made significant strides in text and document translation, but we identified a gap in real-time voice translation solutions," he stated.
Kutylowski highlighted the intricate challenges of creating a real-time translation system, particularly in balancing latency-- the time delay between speech input and translated audio output-- with the need for precision. DeepL is also rolling out add-ons for platforms like Zoom and Microsoft Teams, allowing users to hear real-time translations or view translated text during conversations. Currently, this program is in early access, with organizations invited to join a waitlist.
Furthermore, DeepL's technology facilitates group conversations in settings such as training sessions, enabling participants to join via QR codes. The voice-to-voice system is designed to learn and adapt to specific vocabulary, accommodating industry jargon and personal names.
According to Kutylowski, advancements in AI are reshaping customer service dynamics. He emphasized that a translation layer can empower companies to offer support in languages where skilled personnel are limited and costly to recruit.
DeepL maintains control over its entire voice-to-voice translation process. The current model converts spoken language to text, translates it, and then converts it back to speech. However, the company's future vision includes developing an end-to-end voice translation model that bypasses the text stage entirely, enhancing efficiency and speed.
Despite its innovative approach, DeepL faces competition from various well-funded startups in the translation sector. For instance, Sanas, which recently secured $65 million in funding, employs AI to modify accents in real-time, primarily targeting call center environments. Similarly, Dubai-based Camb.AI specializes in speech synthesis and translation for media companies, aiding in the dubbing and localization of video content.
Another competitor, Palabra, backed by Reddit co-founder Alexis Ohanian's venture firm, is developing a real-time speech translation engine that aims to preserve both meaning and the speaker's original voice, positioning it as a direct rival to DeepL's offerings.
This leap into voice translation not only enhances communication across languages but also represents a significant shift in how businesses can operate globally, fostering greater inclusivity and collaboration in diverse environments.