Announcement

Collapse
No announcement yet.

Google Announces Lyra V2 Low Bit-Rate Voice Codec

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Google Announces Lyra V2 Low Bit-Rate Voice Codec

    Phoronix: Google Announces Lyra V2 Low Bit-Rate Voice Codec

    Last year Google announced the Lyra voice codec for low bit-rates that combined with the open AV1 codec could lead to voice chats on 56kbps connections. Lyra makes use of machine learning and other techniques for extremely low bit-rate speech compression that can function at 3kbps. Google last year open-sourced the Lyra code while today they announced the availability of Lyra V2...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    The SoundStream version showcased last year was able to handle music as well, wonder if Lyra V2 has better success with that. Someone on Discord needs to re-encode 8MB Shrek....

    Comment


    • #3
      3.2kbit/sec that would enable voicechat over things like LoraWAN / could have applications for long range voice transmission via RF with very low transmit powers, as the symbol rate could be very low / the receiver could differenitate the data from background noise via long integration times.

      Comment


      • #4
        The competition in this bit rate for speech isn't Opus (which does fine, but doesn't really work to compete below 8Kb/s), but Codec2. Pick the right target for comparison if you want your results to be meaningful, Google. Now, rate how much computation it takes for the competing codecs and you'll see why Codec2 is used.

        Comment


        • #5
          Originally posted by willmore View Post
          The competition in this bit rate for speech isn't Opus (which does fine, but doesn't really work to compete below 8Kb/s), but Codec2. Pick the right target for comparison if you want your results to be meaningful, Google. Now, rate how much computation it takes for the competing codecs and you'll see why Codec2 is used.
          it would be interesting to see comparisons between Lyra V2 and Codec 2, considering that the minimum Lyra V2 bitrate is the same as the maximum Codec 2 bitrate. Codec 2 works all the way down to 450bps.

          Comment


          • #6
            Originally posted by hotaru View Post

            it would be interesting to see comparisons between Lyra V2 and Codec 2, considering that the minimum Lyra V2 bitrate is the same as the maximum Codec 2 bitrate. Codec 2 works all the way down to 450bps.
            You did read my mind. Google tries to be relevant again, yet it fails AGAIN.

            False propaganda, false innovation.

            Codec2 >>>>>>>>>>>>>>>>>>>>>>>>>> Lyra V2.

            I knew about Codec2 long time ago. I'm interested in radio stuff since many years.
            Last edited by timofonic; 30 September 2022, 06:02 PM.

            Comment


            • #7
              Originally posted by willmore View Post
              The competition in this bit rate for speech isn't Opus (which does fine, but doesn't really work to compete below 8Kb/s), but Codec2. Pick the right target for comparison if you want your results to be meaningful, Google. Now, rate how much computation it takes for the competing codecs and you'll see why Codec2 is used.
              Google's engineers have a tendency to overthink problems. Sometimes coming at problems from a different point of view results in improvements that others can use, so long as Google is willing to give up controlling interest. Other times they tend to impose solutions that are more Rube Goldberg contraptions. Using ML to approach the problem is more of a Rube Goldberg solution since low bandwidth is also more associated with low powered or low resource computing as well. It doesn't match the computing domain really well. Like spoiled rich kid is trying to solve the problem of farming in Africa by buying everyone in a remote village diesel powered tractors. They can't afford or have no access to diesel, so the tractors rot and rust on the side of the field while they go back to using pack animals and elbow grease.

              Comment


              • #8
                Some people may say the audio quality of Lyra V2 may be better at same bitrate, but I think wisely attaching Machine Learning to Codec2 might outperform Lyra V2 in every way.

                A quick 2018 experiment showed very impressive results:https://www.rowetel.com/wordpress/mini.html/?p=5966

                Time for Codec3?

                Comment


                • #9
                  Last year Google announced the Lyra voice codec for low bit-rates that combined with the open AV1 codec could lead to voice chats on 56kbps connections
                  AV1 is a video codec. It is not going to give you voice chats on anything.

                  Comment


                  • #10
                    I assume Michael meant video chat at 56kbps, which is possible with AV1 and Lyra combination.

                    The main competition for Lyra is MS Satin. All Google is trying to do is to keep Google Hangouts competitive with Teams.

                    ML is not running during Lyra encoding or decoding process. Lyra is using a pre-trained neural net. It is roughly as CPU efficient as traditional LPC codecs.

                    Comment

                    Working...
                    X