The new generative AI makes anyone speak a foreign language!

Meta announced Voicebox, an exceptional multilingual text-to-speech AI. Voicebox is a generative model that can generate speech in six different languages from text. But what truly sets it apart is its ability to perform tasks beyond its training, such as editing, noise removal, and even speaker style transfer.

The level of quality it offers is truly outstanding. Meta AI has even decided not to release the Voicebox model or code to the public just yet to prevent any potential misuse.

However, Voicebox does have its limitations. Since it was trained on audiobook data, it may not work as well for casual conversational speech that includes non-verbal sounds. Meta works to remove these issues.

#Meta ⚡️ #TextToSpeach ⚡️ #AI

Meta has shared their Audiocraft neural network — it can create music based on a text description or a test segment 🎶

It works impressively, expect an inflow of beatmakers with samples from classical music.

#Meta ⚡️ #Audio ⚡️ #NeuralNetwork

An AI Alliance has been created to accelerate the development of neural networks.

More than 50 corporations around the world, including Meta, NASA, Intel and Stability AI have joined an alliance.

The organization will combine the computing power and tools of corporations to create open source AI.

This will allow developers to experiment with AI models, improve their security and generally accelerate the development of the field.

The AI Alliance has about 100,000 employees and an annual funding turnover of $80 billion.

Interestingly, Microsoft and OpenAI did not enter there. Apparently, they decided not to help their competitors.

#AIAlliance ⚡️ #Meta ⚡️ #AInews