12 comments on “RED: Improving Audio Quality with Redundancy

  1. Excellent and very relevant post about improving audio quality in webrtc video calls. Great work digging into Zoom and comparing packet lenghts. It looks like we might have to wait for some time until browsers support RED by default. What other approaches do you suggest to improve audio quality with webrtc as of right now?

  2. Probably, SFU should not retransmit RED packets at all, given that the network conditions are so bad (like 60% loss) – to avoid extra congestion and loss.

    • That depends on the subscriber connection, the bandwidth and loss available there. In some cases RED might not be appropriate.

      SFUs are a bit more flexible than browsers though 🙂

  3. Thanks for sharing this, question about the test result, when comparing RED with opus FEC on, is that the opus FEC on the same with RED distance 1, except the RED use the full copy while opus FEC use LBRR? so the concealment should not make much difference? Or the it’s actually “RED distance 1 + OPUS FEC on”, which means it’s actually distance 2?

  4. One more thing, wondering why not use LBRR in RED, is that due to implementation effort or some other technical concerns?

  5. Great post. Thanks for sharing.
    In my wireshark dump, wireshark is not able to decode the packets, i think its due to end to end encryption of zoom.
    Can you provide more detail on setup to dump raw silk frames from zoom.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.