On 1/9/2014 8:29 AM, [email protected] wrote:
I am developing a multi peer audio conferencing software (of sorts) and since using the
Web Audio API handles playback better (faster) than an HTML 5 audio element. I would like
to use it. Firefox is currently supporting Web Audio (AudioContext) and so I am working
on migrating my software over to playing back both local and remote streams using
AudioContext for Firefox (because Chrome does not support
"createMediaStreamSource" on remote streams... yet)
Additionally, for those browsers that support WebRTC and Web Audio we give them
the possibility to use effects (such as distortion and reverb) on their stream.
Here is where I am running into problems and I can't find exactly where the
problem lies... so I'm having difficulty in coming up with a solution.
If a peer using Chrome(31) sends a stream that has effects applied to it using
Web Audio both Firefox(26-29) and Chrome can playback the stream. If a peer is
using Firefox and applies effects using Web Audio neither can playback the
stream.
There is no apparent difference in SDP offers/answers. The only real difference, that I
discovered, is that the Firefox stream taken directly from getUserMedia is labeled as
[object LocalMediaStream] and if I pass it through AudioContext it is then labeled as
[object MediaStream]... even after using "addStream" to set it as my local
stream when setting up the peer connection.
To reduce latency and avoid latency buildup in clock rate mismatch cases
between the internal streams and the microphone input, audio going from
a getUserMedia MediaStream (LocalMediaStream) to a PeerConnection
'bypasses' the MediaStreamGraph itself (PeerConnection uses direct
notification of data being added to the stream as its source, instead of
the processed output of the stream). This needs to be fixed for cases
where the stream added to the PeerConnection isn't a LocalMediaStream,
though that also means that there's a possibility that microphone
sources processed with WebAudio will drift (more delay or occasional
underflows/glitches), which also needs to be resolved with resampling
(bug 908834 and 818822)
I filed a bug to track this, bug 958090. Note that while it can be
fixed on it's own, as mentioned it will cause problems when used until
the resampling bugs are fixed.
--
Randell Jesup, Mozilla
_______________________________________________
dev-media mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-media
