With its new voice model, DeepL Voice now translates intonation and speaking speed into other languages as well
Reported by 파이낸셜뉴스 and The Decoder2 sources ↓
DeepL says its Aure model carries a speaker’s tone, emotion, emphasis and pace into translated speech. It analyzes the original audio and translated text continuously as a conversation unfolds.
Key points
- Aure runs automatically in DeepL Voice for Online Meetings, In-Person Conversations and the DeepL API
- The model supports English, German, Spanish, French, Italian, Japanese, Portuguese and Korean
- Aure replaces a system that used a static voice sample captured at the start of a conversation
Sources
Details
The Decoder reports that DeepL’s Aure model continuously analyzes the speaker’s original audio and the translated text, rather than relying on a voice sample captured at the start of a conversation. DeepL says this lets the model carry tone, emotion, emphasis and speaking pace into translated speech, including as a speaker’s mood changes.
DeepL says Aure runs automatically in DeepL Voice for Online Meetings, In-Person Conversations and the DeepL API. At launch, it supports English, German, Spanish, French, Italian, Japanese, Portuguese and Korean.
According to The Decoder, DeepL separates voice identity, speech melody and content during training so each can be transferred independently. The company says reinforcement learning is used to keep accents out of the output.
Written from coverage as of