Sakhanda Wire
NVDA $153.93 -1.41% MSFT $506.53 -2.21% GOOGL $280.96 -0.97% META $620.53 -0.94% AMZN $256.87 -0.34%
← Back to the news

DeepL introduces voice model for more expressive real-time translation

DeepL is introducing Aure, a model intended to make real-time translations preserve more of the original speaker’s delivery. The company says it transfers tone, emotion, emphasis and speaking pace into the target language, helping translated speech reflect changes in mood and elements such as sarcasm.

Aure continuously examines the source audio and the translated text during a conversation. During training, DeepL separates voice identity, speech melody and linguistic content, allowing each part to be handled independently in another language. The company says reinforcement learning is used to prevent accents from entering the generated output.

The model is available 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. DeepL introduced voice preservation in September. The company has also acquired the team behind Mixhalo and is restructuring around smaller AI-focused teams after cutting about 250 jobs in May. Separately, Block has presented Dessa, a speech model designed to identify emotions and sarcasm from audio.

Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.

Originally published by The Decoder on

Read the original on The Decoder ↗

The reporting and the image belong to The Decoder and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news