DeepL develops Aure model to carry tone and mood into real-time voice translation
With the model Aure, the company behind DeepL aims to make stress, speaking pace and mood audible in translation as well.
The company DeepL is working on a new generation of its real-time translation. With the model named Aure, DeepL Voice is meant to carry not only words but also tone and emotional state into the target language. According to the available information, the goal is for spoken translations to sound more natural.
According to the available information, Aure transports, alongside emotion, the stress and speaking pace of the original into the translated speech. The aim is for listeners to experience the translation in real time as if the person were speaking themselves, with their urgency, restraint or irony. The headline under which the topic was circulated presents this goal in a particularly pointed way: sarcasm and changes of mood are supposed to be captured reliably for the first time.
How well this actually works is not evident from the information known so far. Neither benchmarks nor accuracy figures were mentioned that would show how reliably sarcasm and changes of mood are recognized and conveyed. The announcement thus remains within the realm of a claim, without measurable evidence in the text itself.
It also remains unclear whether the feature is already available or still in development. No release date was mentioned, nor any details about supported languages or pricing. As long as these details are missing, it can only be stated that DeepL has announced a model that is intended to preserve the emotional content of spoken language in real-time translations.
Source: the-decoder.com
16
