[vlc-devel] [PATCH 1/1] opus: handle RTP depayloading (fixes #11938)

[Tristan Matthews]

On Sun, Jan 11, 2015 at 11:25 AM, Ross Finlayson <finlayson at live555.com> wrote:
>
> On Jan 11, 2015, at 6:24 AM, Rémi Denis-Courmont <remi at remlab.net> wrote:
>
> Le 2015-01-10 01:55, Tristan Matthews a écrit :
>
> If no header is given, deduce it from the decoder format.
>
>
> There is always a clock rate in RTP. You get it in the SDP rtpmap line.
>
>
> Note that this is already done in line 857 of "live555.cpp", where you do
> (for audio tracks):
> tk->fmt.audio.i_rate = sub->rtpTimestampFrequency();
> to set the *default* audio sampling frequency.  So you shouldn't need to do
> this again.

It wasn't being done twice, only in the case that it was set to 0,
which apparently can't happen so this case will be dropped.

>
> In any case, however, note the Opus RTP payload format document:
> https://tools.ietf.org/html/draft-ietf-payload-rtp-opus-06
>
> In particular:
> 1/ For Opus, the RTP clock rate is always 48000 Hz
> 2/ HOWEVER, the sampling frequency of the Opus audio that's contained within
> the RTP packet can be 8000 Hz, 12000 Hz, 16000 Hz, 24000 Hz, or 48000 Hz;
> see "Table 1".  This, not the RTP clock rate, is what you care about for
> VLC.  (Only the LIVE555 library should really care about the RTP clock
> rate.)
>
> Note also, in section 4.2:
>
>    The Opus encoder can output encoded frames representing 2.5, 5, 10,
>    20, 40, or 60 ms of speech or audio data.  Further, an arbitrary
>    number of frames can be combined into a packet, up to a maximum
>    packet duration representing 120 ms of speech or audio data.  The
>    grouping of one or more Opus frames into a single Opus packet is
>    defined in Section 3 of [RFC6716].  An RTP payload MUST contain
>    exactly one Opus packet as defined by that document.
>
> Note that section 3.1 of RFC 6716
> https://tools.ietf.org/html/rfc6716#section-3
> describes how to deduce the sampling frequency and number of frames from the
> first byte of each "Opus packet" - i.e., the data that's contained in each
> Opus RTP packet.  I hope, however, that you already do this elsewhere in
> VLC, in its generic code for decoding Opus audio.  You should not have to be
> doing this (looking at the first byte of payload data to deduce sampling
> frequency, etc.) in the "live555.cpp" module, as that code should be
> specifically for RTSP/RTP-related stuff.

Thanks for the info, I also tested the patch for sample rates of 8000,
12000, 16000, 24000 and 48000 and had no issues (with both our libopus
decoder and the libavcodec opus decoder).

Best,
Tristan