Google Posts Initial Code For Lyra Speech Codec

[phoronix]

Phoronix: Google Posts Initial Code For Lyra Speech Codec

Back in February we covered Google's work on the Lyra voice/audio codec designed for fitting with very low bit-rate audio for speech compression in use-cases like WebRTC and video chatting even on the most limited Internet connections. Thanks to leveraging machine learning, Lyra can function at just 3kbps. The code to Lyra is now public...

https://www.phoronix.com/scan.php?page=news_item&px=Google-Lyra-Open-Source

[ThoreauHD]

Why does this level of video compression make me think of no such agency. It's not for the third world. They have latency and error correction issues. It's not for the Moon, because we don't go there anymore. So... Cui bono?

[BesiegedAce]

Why does this level of video compression make me think of no such agency. It's not for the third world. They have latency and error correction issues. It's not for the Moon, because we don't go there anymore. So... Cui bono?

"voice/audio codec designed for fitting with very low bit-rate audio"
'video compression'

Sometimes a cigar is just a cigar. Should we instead be designing extremely inefficient and bulky codecs instead?

[Min1123]

I am not getting along with Google's build system: bazel, which is needed to compile this. Feels like somebody wanted a non-Java maven, and this got slapped together. The current Lyra source requires a specific checkout of clang and is Ubuntu-aligned for building. I'm going to go use podman with Ubuntu 20.04 on my Fedora 34 to try to play with this.

[Setif]

Fix it for you:

It's not for the Moon, because we didn't go there. So... Cui bono?

[bemerk]

How well does it work with female voices? Studies show that they are currently discrimated against in Web conferencing because high frequencies are often cut out in the transmission and females therefore have a lower chance to express their emotions properly

[Alex/AT]

"Open source" is a bit of stretch there. It is based on closed-source library which source code is "impossible to open".

[archkde]

"Open source" is a bit of stretch there. It is based on closed-source library which source code is "impossible to open".

Is this blob also governed by the Apache License, permitting reverse engineering? But unfortunately, it is quite big.

[oibaf]

From the github project page:

Please note that there is a closed-source kernel used for math operations that is linked via a shared object called libsparse_inference.so. We provide the libsparse_inference.so library to be linked, but are unable to provide source for it. This is the reason that a specific toolchain/compiler is required.

[lucasbekker]

From the github project page:

Please note that there is a closed-source kernel used for math operations that is linked via a shared object called libsparse_inference.so. We provide the libsparse_inference.so library to be linked, but are unable to provide source for it. This is the reason that a specific toolchain/compiler is required.

If it is just a single kernel, it shouldn't be too hard to replace it with some calls to openBLAS or something like that, won't be as fast though...