• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Meta’s AI-powered audio codec EnCodec promises 10x compression over MP3

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
Meta (Facebook owner) announced an AI-powered audio compression method called "EnCodec" that can reportedly compress audio 10x smaller than the MP3 format at 64kbps with no loss in quality. Meta says this technique could dramatically improve the sound quality of music and speech on low-bandwidth connections, such as phone calls in areas with spotty service.

Meta describes its method as a three-part system trained to compress audio to a desired target size. First, the encoder transforms uncompressed data into a lower frame rate latent space representation. The quantiser then compresses the representation to the target size while keeping track of the most important information that will later be used to rebuild the original signal. This compressed signal is what gets sent through a network or saved to disk. Finally, the decoder turns the compressed data back into audio in real time using a neural network on a single CPU.

Their post is here and the paper here.

I assume this will be important for developing markets and rural USA (I couldn’t resist the tease) but not for the developed countries where fibre with Giga speeds will soon reach saturation levels. In the UK almost all new builds have fibre.
 

dorakeg

Senior Member
Joined
Jul 20, 2022
Messages
326
Likes
187
Meta (Facebook owner) announced an AI-powered audio compression method called "EnCodec" that can reportedly compress audio 10x smaller than the MP3 format at 64kbps with no loss in quality. Meta says this technique could dramatically improve the sound quality of music and speech on low-bandwidth connections, such as phone calls in areas with spotty service.

Meta describes its method as a three-part system trained to compress audio to a desired target size. First, the encoder transforms uncompressed data into a lower frame rate latent space representation. The quantiser then compresses the representation to the target size while keeping track of the most important information that will later be used to rebuild the original signal. This compressed signal is what gets sent through a network or saved to disk. Finally, the decoder turns the compressed data back into audio in real time using a neural network on a single CPU.

Their post is here and the paper here.

I assume this will be important for developing markets and rural USA (I couldn’t resist the tease) but not for the developed countries where fibre with Giga speeds will soon reach saturation levels. In the UK almost all new builds have fibre.

I would say it's even more important for urban than rural. Mobile users and ISPs will greatly appreciate this features. Esp. in crowded places.

Less data downloaded means lower traffic, less congestion. Then, many of us are still on mobile plan where we are charged by amount of data (eg. 50GB per month).

The only issue would be power consumption and heat. Does this feature uses a lot more power?
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
but not for the developed countries where fibre with Giga speeds will soon reach saturation levels. In the UK almost all new builds have fibre.
Some countries have unlimited mobile data plans for reasonable cost but majority don't.

So still good news for mobile streaming.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
The Ground Truth (original) is good enough for me. Their AI compressed audio just sounds horrible.
 
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
The Ground Truth (original) is good enough for me. Their AI compressed audio just sounds horrible.
EnCodec is never meant to be a Hi-Fi codec, but you cannot be unimpressed with the quality of the sound at just 6kbps. It is an engineering marvel!

The world listens music on cars, kitchen radios, TVs or BT speakers. What quality loss you will hear when enCodec is used?
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,222
Likes
17,799
Location
Netherlands
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
You guys are late to the party:

I missed that post and created this one because they named the codec wrong. It is not called SOTA as the title says. That name is not even mentioned on the linked post..
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,222
Likes
17,799
Location
Netherlands
I missed that post and created this one because they named the codec wrong. It is not called SOTA as the title says. That name is not even mentioned on the linked post..
No worries. I have no idea where the SOTA name comes from... don't blame me ;)
 
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
I had some trouble finding the other topic again, so I am not surprised this one popped up.
I guess that's because neither the publisher's nor the codec's name was on that post.
 
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
OP
sarumbear

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
SOTA = State Of The Art.

Not a name, but rather a claim.
Well, as the title says: "New AI compression algorithm called SOTA" it clearly means a name, not a claim, hence the confusion.
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,222
Likes
17,799
Location
Netherlands
I understand the confusion, but also understand what that OP meant. In this case, "called" is used in the sense that "some are calling it...".
How about:
New SOTA AI compression algorithm from Meta claims 10x better compression than MP3
 
Top Bottom