Google Lyra codec

[w-u-2-o]

So what do people think about the Google Lyra codec? As of 6 Apr it's been released as open source:

https://opensource.googleblog.com/2021/04/lyra-enabling-voice-calls-for-next-billion-users.html

It is demonstrating some really spectacular performance at a 3kbit/s data rate, far better than MELP or any other choice at that data rate IMHO. Examples here:

https://ai.googleblog.com/2021/02/lyra-new-very-low-bitrate-codec-for.html

[Bruce Perens]

It includes a proprietary binary module. I don't know what that does.

--
You received this message because you are subscribed to the Google Groups "digitalvoice" group.
To unsubscribe from this group and stop receiving emails from it, send an email to digitalvoice...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/digitalvoice/ec2edf6b-65a7-44d4-82a6-ee7ab8866341n%40googlegroups.com.

[w-u-2-o]

The "proprietary binary module" may be that as described in the article as "a math kernel, for which a shared library is provided until we can implement a fully open solution over more platforms".

So if that is correct, and if we believe Google, it should be 100% open source soon.

[Stuart Longland]

On 8/4/21 1:09 am, w-u-2-o wrote:
> So what do people think about the Google Lyra codec? As of 6 Apr it's been
> released as open source:

It was discussed on the Codec2 list:
https://sourceforge.net/p/freetel/mailman/freetel-codec2/thread/BYAPR18MB2518F052F3B21817D33B3EF9A09C9%40BYAPR18MB2518.namprd18.prod.outlook.com/#msg37229049

Probably too high a bit rate to use for FreeDV, but might work in
higher-bitrate applications.
--
Stuart Longland (aka Redhatter, VK4MSL)

I haven't lost my mind...
...it's backed up on a tape somewhere.

[Rafael Diniz]

As you say, there is a .so compiled with a specific version o clang, and
you also need that specific version of clang to compile the codec.

Also, it will need to be way better than LPCNet, as the bitrate is twice
the LCPNet bitrate. In the end, at 3 kbit/s is hard to fit in a ~2.4 kHz
bandwidth of a SSB HF narrowband radio.

Rafael
PU2UIT

> <https://groups.google.com/d/msgid/digitalvoice/ec2edf6b-65a7-44d4-82a6-ee7ab8866341n%40googlegroups.com?utm_medium=email&utm_source=footer>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "digitalvoice" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to digitalvoice...@googlegroups.com

> <mailto:digitalvoice...@googlegroups.com>.

> To view this discussion on the web visit

> https://groups.google.com/d/msgid/digitalvoice/e2723fbf-f68b-4088-b6dd-ee15697385a8n%40googlegroups.com
> <https://groups.google.com/d/msgid/digitalvoice/e2723fbf-f68b-4088-b6dd-ee15697385a8n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[David Rowe]

Thanks Stuart.  Yes at 3000 bits/s it is useful in VHF/UHF (and perhaps
satellite) applications. Given it's open source, it could be modified to
operate at a lower bit rate (and quality) to squeeze through a HF
channel at a reasonable SNR.

- David

[Adrian Musceac]

What I would be interested in is a comparison with Codec2 at a close bitrate, using longer and varied audio samples from multiple speakers.
I see no reason why Codec2 at 3 - 3.2 kbps could not perform just as well but with less CPU power.

Adrian

[Bruce Perens]

I am most glad it is Open Source because of Google's patent. There was a chance that they might have assertred it, before.

--

You received this message because you are subscribed to the Google Groups "digitalvoice" group.
To unsubscribe from this group and stop receiving emails from it, send an email to digitalvoice...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/digitalvoice/55A0276A-ADAF-4035-86AB-020F9C699F5C%40gmail.com.