As you know, the main Tech players have entered the AI race. After the launch of ChatGPT at the end of 2022 and Microsoft’s $10 billion investment in the OpenAI initiative, web giants rushed to deliver their own AI in turn.
Google came forward with its conversational AI Bard, while Meta confirmed in April 2023 that its AI was in development. In recent months, the Menlo Park firm has released a number of artificial intelligence models, starting with LLaMA (Large Language Model Meta Artificial Intelligence). an open source language model.
A while ago, the Californian company also announced JEPAA model that aims to reproduce human thought, especially by analyzing and understanding abstract concepts and concepts. In a completely different area Meta also presented MusicGenAn artificial intelligence that can create music through a basic text description.
Meta introduces Voicebox, AI that can mimic the human voice
But on June 16, 2023, Meta said, “his new breakthrough in generative artificial intelligence for speech”. This artificial intelligence is Voicebox. In short, this cutting-edge AI model specializes in: voice synthesis. In other words, he is able to create, edit or format audio files.
First, let’s address Voicebox’s most interesting (and probably most problematic) feature: text-to-speech synthesis in context. Based on just a two-second excerpt, Voicebox can generate a speech like this: simulating the voice and phrases of the person heard in the quote.
In this way, Voicebox can simulate the voice of a relative, a singer or a politician. Meta says that in the future, Voicebox and other similar productive AI models will be able to: give voice assistants natural sounds or NPCs in the metadata store. They can also enable the visually impaired to hear messages written by their friends’ voices.
Also to read : After Dall-E and Midjourney, this new AI can create video from a text
Editing of audio files and instant translation
But that’s not all, because Voicebox also offers other features:
- Audio editing and noise reduction : Voicebox can recreate a part of speech that was interrupted by noise, or replace slurred and mispronounced words without the need to re-record the entire speech (a kind of Google-like magic eraser for audio)
- multilingual translation : Voicebox currently supports six languages (English, French, Spanish, German, Polish, and Portuguese), allowing it to export the speech to a different language than the original file (while importing styles and shadows)
Meta’s AI has been perfected in multiple areas to perform its various tasks. 50,000 hours extract sound mostly from audiobooks and royalty-free content. for now The soundbox is inaccessible to the general public, for security reasons. Not surprisingly, Meta is concerned about the abuse of its artificial intelligence, including imitating the voices of real people.
Source : Meta
