Goal continues in the construction of its metaverse and for that it needs to create a functional environment in a scenario that can be changed according to the user’s availabilities. One of his recent works is in the consolidation of an audio format much more compressed than MP3.
For almost 30 years, this type of file has established itself as the Internet’s favorite, because it allows high quality without taking up a very large space, which is ideal for unstable browsing or to allow ideal use of the web without loading problems. .
What the company is doing mark zuckerberg is to create an audio hyper-compression codec, which allows the same quality of MP3 but with a much smaller size, ideal for not affecting the navigation of people on different platforms.
new audio format
To carry out this development Goal It is relying on artificial intelligence to train neural networks so that they understand the way in which the audio is built, then it is recreated and reproduced with the same quality, but with less space than the original.
“We achieved a compression rate of about 10 times compared to MP3 at 64 kbps, with no loss of quality. While these techniques have already been explored for speech, we are the first to make them work for stereo audio sampled at 48 kHz (i.e. CD-quality), which is the standard for music distribution.
The operation of this process is carried out in three steps, to take the audio and take it to a smaller format:
– The encoder takes the uncompressed data and transforms it into a higher dimensional, lower frame rate representation.
– The quantifier compresses this representation to the size that we propose. This step is trained to provide the desired size while preserving the most important information to reconstruct the original signal. This compressed representation is what is stored on disk or sent over the network. It is the equivalent of the .mp3 file.
– The decoder is the last step. Here the compressed signal is converted back into a waveform as close as possible to the original.
This entire process allows “compressing and decompressing audio in real time with state-of-the-art size reductions”, with which they seek several objectives, such as creating faster calls and improving conditions when connected to poor networks, in addition to delivering more user-friendly experiences. comfortable in the metaverse without requiring large connections.
for now, Goal He assured that it is focused on audio, so this process through artificial intelligence is not focused on video, but it is the beginning to improve conditions in formats such as video calls, movie streaming and virtual reality video games.
This would also mean that the audio and video chips of telephones and computers would have to be improved to adapt to this technology, which will allow them to consume less energy.
“We want to continue exploring how we can compress audio to even smaller file sizes without significantly degrading quality. We also plan to explore spatial audio compression, which will require a technique that can compress multiple channels of audio while maintaining accurate spatial information. These learnings could be useful for future metaverse experiences.” Goal.