Universal communication, beyond languages, is an old dream of humanity. This artificial intelligence realizes this claim and at the same time reproduces the expression of the voice…
This will also interest you
[EN VIDÉO] Foreign languages are good for memory. Speaking multiple languages has many benefits for the brain and maintains mental state when…
Facebook (Meta) has been working on artificial intelligence for several years and making it known. Mark Zuckerberg’s company even opened a research center in Paris and integrated a dedicated researcher into its ranks: Yann Le Cun. And yet… Like others, Meta was caught off guard by OpenAI, which quickly established itself as a market leader in artificial intelligence with ChatGPT. Since then, Meta appears to have been slow to demonstrate his prowess, although breakthroughs were made, particularly in musical creation or in the analysis of mental images. However, this research appeared to have difficulty being successful.
Meta could definitely make history with Seamless…
Like in a science fiction film
Imagine the scene: you are in Peru, Shanghai or Gothenburg. You speak French normally and the person you are speaking to (who is wearing headphones) receives your words in their native language, in the form of expression you use.
That’s what Seamless achieves; something we usually see in science fiction films or novels: the instant translation of conversations. In over a hundred languages!
Two second delay
In reality, Seamless achieves this with a delay of two seconds – and even slightly less, if the official speech is to be believed. And there’s a bonus: the tone of voice is reproduced. Depending on whether you are whispering, expressing sadness or anger, Seamless tries to reproduce the same impression. In short, your interviewer will pick up on the language you used.
A demo accessible online
The version of Seamless that is used in everyday life is not yet available – but is planned for 2024. However, a demo of Expressive mode is accessible at this address: Seamless Expressive. You need to activate your webcam and microphone and then express yourself in the mode of your choice. Upon arrival you will receive your video in English, Spanish or German. We ran various tests and what can we say… Seamless reproduced our expressions quite well, although progress is still possible. You can even have fun sending texts at full speed, and you’ll find that Seamless is up to the challenge. Note that the tool also allows you to listen to the “expressionless” representation to better assess its performance. And obviously the difference is remarkable!