Announcement

Collapse
No announcement yet.

Google's Lyra v0.0.2 Speech Codec Gets Rid Of The Binary Blob

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

  • Google's Lyra v0.0.2 Speech Codec Gets Rid Of The Binary Blob

    Phoronix: Google's Lyra v0.0.2 Speech Codec Gets Rid Of The Binary Blob

    Earlier this year Google announced the Lyra voice codec that could work with AV1 video for video chats over 56kbps modems. Google is today shipping its newest Lyra version...

    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
    It's still looking promising, has anyone tried it?

    Comment


    • #3
      I'm curious about it as well as it would be a direct competitor with Opus codec which I have used and is quite good. Google has a lot of NIH syndrome so it makes me wary of anything they release, especially with their penchant of abandoning projects and products regardless of popularity.

      Comment


      • #4
        Originally posted by stormcrow View Post
        I'm curious about it as well as it would be a direct competitor with Opus codec which I have used and is quite good. Google has a lot of NIH syndrome so it makes me wary of anything they release, especially with their penchant of abandoning projects and products regardless of popularity.
        Funny you mention that. My first curiosity was if Opus and other codecs could make use of the sparse matrix multiplication library to get better low bitrate audio (and if that was possible if it could be used to shave some kbps off of higher quality audio settings).

        Comment


        • #5
          Originally posted by skeevy420 View Post

          Funny you mention that. My first curiosity was if Opus and other codecs could make use of the sparse matrix multiplication library to get better low bitrate audio (and if that was possible if it could be used to shave some kbps off of higher quality audio settings).
          The special sparse MM impl is for performance, it's not a coding efficiency tool.

          Comment


          • #6
            Originally posted by stormcrow View Post
            I'm curious about it as well as it would be a direct competitor with Opus codec which I have used and is quite good.
            Opus also comes with a license that is revocable in the case of certain types of lawsuits. Some consider that a restrictive license, and Facebook tried a similar license and was banned from contributing from certain communities until they changed the license of their sources to be MIT. Lyra's license is Apache-2.0.

            Comment


            • #7
              Originally posted by CommunityMember View Post

              Opus also comes with a license that is revocable in the case of certain types of lawsuits. Some consider that a restrictive license, and Facebook tried a similar license and was banned from contributing from certain communities until they changed the license of their sources to be MIT. Lyra's license is Apache-2.0.
              I think you mean that the Opus patent license terminates in case of a patent suit. The Apache license does that as well, so there's no difference in this regard.

              Comment


              • #8
                still liking the look of this for mode3 LoraWan...

                Comment


                • #9
                  Originally posted by archkde View Post

                  I think you mean that the Opus patent license terminates in case of a patent suit. The Apache license does that as well, so there's no difference in this regard.
                  My comment wasn't aimed at the licensing issues, only at the merits of the codecs and Google's well documented problems with focus - potentially leading to project abandonment in the future.

                  Comment


                  • #10
                    Tried it with episodes of some podcasts I listen to and was rather disappointed. While I was well prepared to encounter distorted voices, audible artifacts, and perhaps even some loss in legibility (especially during crosstalks) the amount of information that my brain failed to decode in time was simply too high for me to enjoy the experience. In comparison, the same content encoded in 24kbps opus retained its quality exactly as I perceived it with the original. So as far as this application is concerned, the 8-fold reduction in size is in my view not worth it. But was Google ever to enable a way for users of Lyra to refine the models further by training on their own datasets, I'd be interested in testing it out.

                    Comment

                    Working...
                    X