With high compression, the technology offered by Meta can enable high quality calls and music even with poor connections.
Photo: Harbucks – shutterstock.com
Facebook and WhatsApp’s Meta company released an AI-powered audio compression method called “EnCodec” last week announce. According to the report, the technology can compress audio data ten times less than the MP3 format at 64 kbps without sacrificing quality. According to Meta, this technology can improve voice quality over low-bandwidth connections, for example B. With phone calls in areas with intermittent service, it is significantly improved. The technology also works with music and thus could replace MP3 in the future. Meta AI showed the difference between different compression techniques With a valid example.
Meta presented the technology on October 25 in a paper titled High-Resolution Neural Compression It was written by Meta AI researchers Alexandre DeVosses, Jed Kobet, Gabriel Senevi, and Yossi Ade. Meta also summarized the results of the research on his EnCodec blog.
Photo: MetaAI
Meta describes its method as a three-part system that has been trained to compress sound to a desired target volume. First, the encoder converts the uncompressed data into a lower representation of the frame rate. The quantifier then compresses the representation to the size of the target, tracking the most important information that is later used to reconstruct the original signal. This compressed signal is sent over a network or stored on disk. Finally, the decoder converts the compressed data into one using a neural network CPU Return to audio data in real time.
Photo: MetaAI
The key to developing a way to compress audio data as much as possible without losing key elements of the signal is to use Advantages. Meta explains that these have the task of distinguishing between real and reconstructed samples. The compression model attempts to create samples to deceive the discriminators by making the reconstructed samples appear more perceptually similar to the original samples.
As far as the application domain is concerned, Meta primarily sees the potential to use AI-powered “voice over-compression” to enable “faster calls with better quality” in poor network conditions. Unsurprisingly, the researchers also point out that the technology is finallyRich metaverse experiences Without major improvements in bandwidth.
“Certified tv guru. Reader. Professional writer. Avid introvert. Extreme pop culture buff.”
More Stories
Remotely controlled cargo ships coming soon on the Elbe Canal?
Siemens technology makes Baden Canton Hospital smart
Discovering an ancient Mayan city – what do the rainforests hide?