voice box
“Risky for the Audience” – Facebook’s new artificial intelligence perfectly mimics voices
With just two seconds of sampled voice, Voicebox, Meta’s new artificial intelligence, can perfectly simulate a person’s voice. However, the AI model remains classified for ethical reasons.
updated
Voice Generative AI sparked controversy in the past when rapper Jay-Z’s voice was used for a new song he never sang.
20 m / Tariq Al-Sayed
-
Meta (Facebook) has announced a new voice generation app.
-
Using artificial intelligence, almost perfect voice recordings of people are created.
-
They are generated using text-to-speech or existing audio files.
-
The tool aims to revolutionize different types of communication in the future.
Will we soon stop speaking podcasts, audio news, and foreign languages ourselves? a New AI model Powered by Facebook Meta simulates and generates sounds in a near-perfect human way, including pauses and patterns. Voice box was from AI researchers at Meta It was developed as the first major model focused on speech generation.
This is how it works
Similar to the AI that generates text or images, whether it is ChatGPT or Midjourney, Voicebox can also generate multiple outputs based on text or voice. The model can synthesize sounds in six languages. Spoken content can be changed and noise or interruptions removed. It is also possible to use your voice to create perfectly spoken sentences in other languages. The AI has been trained with over 50,000 hours of spoken audio files.
These are two examples of how sounds are generated by text in Voicebox. Existing audio samples can also specify what is spoken.
meta
Voicebox only takes 2 seconds of audio samples to perfectly imitate your voice. Meta wants to build the app for various application domains, including:
-
Speech generation for people who have lost their voice due to surgery, illness, or other reasons.
-
Make language translation accessible and empower people to speak foreign languages with their own voice.
-
As a productivity tool, Voicebox can edit spoken content, remove noise, create content for businesses, marketing, TV, and more.
Those are the concerns
Will Whatsapp voice messages be generated via voice box soon?
Getty Images/iStockphoto
According to Facebook/Meta, the technology is still very dangerous, but it should be discoverable in the future.
Getty Images/iStockphoto
The developers stress in the letter that they are aware of the ethical consequences and have not yet released the app at this time. “While Voicebox can have many positive social impacts, it also carries the potential for misuse and unintended harm,” she said. Research report authors. This includes deepfakes or fake audio recordings of strangers.
To address this problem, the AI developers at Meta trained the model to accurately distinguish between real speech and artificial speech. They want to incorporate artificial digital “fingerprints” into audio samples that can be easily identified without degrading the sound quality.
Once the security is guaranteed, Meta wants to revolutionize the language and fully communicate with Voicebox in the future and achieve a similar effect as ChatGPT.
With the daily update, you stay up to date on your favorite topics and never miss any more news about current world events.
Receive the most important information directly in your mailbox every day.
“Tv expert. Hardcore creator. Extreme music fan. Lifelong twitter geek. Certified travel enthusiast. Baconaholic. Pop culture nerd. Reader. Freelance student.”
More Stories
People – American star Kim Kardashian in Hamburg – Economy
RICO is interested in the Cilander area – appenzell24.ch
That general meeting was 2024