DeepL launches real-time voice translation preserving speaker nuances

DeepL has launched a new real-time voice translation service, ‘DeepL Voice’, featuring a Voice AI model designed to preserve the speaker’s original voice characteristics and emotional nuances during translation. The technology maintains the speaker's unique tone, tempo, and emotional expressions—such as hesitation, urgency, or excitement—across different languages.

The service supports voice-to-voice translation in over 30 languages, with 14 specific languages, including Japanese, English, Chinese, Spanish, and German, capable of maintaining the speaker’s specific voice profile.

DeepL also introduced desktop applications for Windows and Mac that integrate with major online meeting platforms, including Zoom, Microsoft Teams, and Google Meet. These apps can automatically detect meetings to provide real-time translated subtitles on-screen or allow users to switch to audio translation. According to independent testing by Slator, DeepL Voice achieved a translation error rate of 4%, performing significantly better than the average error rate of 17% reported for the integrated tools in Zoom, Teams, and Google Meet.