Re: [codec] Suggestions for OggOpus draft

[Ralph Giles <[e-mail]>]

On 12-09-24 10:32 AM, Jean-Marc Valin wrote:

> I suggest adding these two sections to the OggOpus draft, one that
> describes how to handle chaining and one that has recommendations for
> encoders. I'm sure there's still more things an encoder can do that I
> haven't listed.

Thanks for writing this up.  A couple of questions:

By 'linear prediction' do you mean the LPC-based lapping the opusenc
utility already uses on the final packet?

Is having the encoder supply predicted audio only to help decoders which
don't do a cross-fade, or does it also help preserve fidelity
across the boundary even in the case that the decoder does something to
mitigate the discontinuity?

I guess the issue is that if the encoder is processing a section of a
larger stream, with the intention that the output will generally be
concatenated, like tracks within an album with continuous audio, it
should pad with the actual earlier and later samples. Prediction should
therefore be used to smooth the transition only if adjacent audio isn't
available? Or does that decision depend on the quality of the adjacent
audio. E.g. a large transient would disturb the transition when the next
segment *isn't* the expected one, in shuffle play or a custom playlist,
for example?

These don't seem like they're practices decoders can rely on, so they
need to handle dicontinuities regardless. I think it is useful to give
some suggestions in the draft, though if we can reduce the emphasis a bit.

 -r