[vlc-devel] Zero-latency VLC

[David Glaude]

If memory serve me well,

Hi did some attempt at Low-Latency VLC is about the same condition except
that my source is not H.264 but plain simple MPEG4 (I think Part 2 rather
than Part 10) and is also a low lattency encoder. Some of my attempt might
have translate into question on this mailing-list and even some ticket on
trak.

Depending on how your RTSP/RTP stream is describe, the effect of the buffer
control parameter might be different. My multicast stream (wich is video
only) was described by some an ASX file but not the recovered by RTSP
DESCRIBE but either a local file or by doing an HTTP GET. Because my input
is video only, there is no sync problem that could increase the lattency.

So you have to find wich input module is used by VLC, in my case it was
LIVE from LIVE555.

>From my logs, It seems that I was using "--rtsp-caching=120" (I have also
seen "--caching=160" is some of my attempt). What I remember is that there
was some lower limit and whatever value I was trying it was not going below
that limit.

Also, I have discovered that in recent version, VLC try to increase the
input buffering when starvation occure... but this can be disabled by using
"--clock-jitter=0"

My main problem was "stability" of the lattency. Overtime and especially if
there were some network disturbence (packet loss), VLC would be less and
less in sync with the source (I need it to work for 4 days in a raw at the
minimum) and it is the (not well supported) vlc plugin I would use in final
version. My benchmark was QT that as some non compressible 3 second of
buffer... but this is stable over time.

Please share the knowledge you have already accumulated.

Anyway, I need to redo some test with lattest version.

David Glaude

2012/7/5 Maurizio Sciglio <maurizio.sciglio at gmail.com>

> Hi Guys,
>
> I'm trying to figure out whether VLC can be used to play a H.264 video
> streamed using RTSP/RTP with (virtually) zero latency. By using
> "network-caching" I managed to reduce the latency to ~80-100ms but if I try
> to reduce it any further the video won't play at all.
>
> I am aware that VLC hasn't been designed with that goal in mind but I
> would greatly appreciate if you could suggest anything I could try or,
> alternatively, areas of the code I could/should look into in order to
> modify VLC for what I need.
>
> Thanks!
> Maurizio
>
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> http://mailman.videolan.org/listinfo/vlc-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20120705/990b4426/attachment.html>