[QLab] Problem playing video over 4 screens

291 views
Skip to first unread message

Geordie Barker

unread,
Feb 5, 2010, 1:21:59 AM2/5/10
to Discussion and support for QLab users.
Our setup is 4 screens of 1024 by 768, run on a Mac Pro 2.93 Ghz Quad Core on 10.6.2. Two dual head ATI Radeon HD 2600 XT video cards feed four projectors via VGA.

I have a video file that is 25 frames a second, resolution 4096 by 600 in the format of H.264.

When I run it in Qlab across the 4 screens (that run side by side), it stutters, drops frames and doesn't play back properly.

When I use MSAQuicktime Player to play it back across 4 screens, the result is flawless.

If I split the file into 4 seperate 1024 by 768 files, it plays back in Qlab with minimal problems.

Any ideas?

Geordie 

Christopher Ashworth

unread,
Feb 5, 2010, 9:48:26 AM2/5/10
to Discussion and support for QLab users.
Hi Geordie,

The way QLab's video engine is constructed, it has an optimized
rendering path when the video is assigned to only one screen.

Basically, when a video goes to multiple screens, the whole video
frame must be decoded, placed into intermediary storage, and then sent
to each separate video card.

In contrast, when a video goes to a single screen, the frame is
rendered directly to that single graphics card.

So in the two cases you describe, the one with a single large video
has a full 4096 x 600 video frame decoded into the intermediary
location and then that full frame is copied 4 times and sent to each
graphics card. So what is ending up happening is that there is a lot
of extra computing going on when using the single widescreen video:
lots of extra copies and lots of extra pixels. (i.e. the extra pixels
are the ones that are never rendered on a screen, since each screen as
the full 4096x600 frame, but only shows part of it.)

In the case of four separate smaller videos, you are in essence
manually optimizing away those extra copies and extra pixels.

One thing you might want to try is to copy the original large video
cue into 4 separate cues, and assign each to only a single screen.
This would be something in between the two scenarios you described:
it would still involve decoding extra pixels, but it would at least
allow QLab to use the more optimal rendering path of each video
directly to the graphics card. It might well give you enough
performance to use this as a shortcut instead of needing to split the
file up.

Obviously the best thing would be for us to find ways to implement
these kinds of optimizations automatically so you don't have to know
or think about any of this stuff. I would hope we could make some
improvements in that area for v3. In the meantime, hopefully this is
helpful.

Best,
Chris

________________________________________________________
WHEN REPLYING, PLEASE QUOTE ONLY WHAT YOU NEED. Thanks!
Change your preferences or unsubscribe here:
http://lists.figure53.com/listinfo.cgi/qlab-figure53.com

Ben Chaisson

unread,
Feb 5, 2010, 11:57:28 AM2/5/10
to Discussion and support for QLab users.
Format format format.
 Quicktime .mov  is not one single format. The best and easiest format i have found is H.264 if i use the Ntsc Dv format everything goes to hell.
use Compressor to batch convert all your files to H.264
Ben Chaisson

--- On Fri, 2/5/10, Geordie Barker <geor...@gmail.com> wrote:
-----Inline Attachment Follows-----


________________________________________________________
WHEN REPLYING, PLEASE QUOTE ONLY WHAT YOU NEED.  Thanks!
Change your preferences or unsubscribe here:
http://lists.figure53.com/listinfo.cgi/qlab-figure53.com


Yahoo! Canada Toolbar : Search from anywhere on the web and bookmark your favourite sites. Download it now!

Andy Dolph

unread,
Feb 5, 2010, 12:13:08 PM2/5/10
to Discussion and support for QLab users.
So Chris, would it be best to divide this file into 4 separate files
with just the portion of the image that's projected on each screen and
trust qlab to run them in sync?

Andy

Christopher Ashworth

unread,
Feb 5, 2010, 2:03:41 PM2/5/10
to Discussion and support for QLab users.
On Feb 5, 2010, at 12:13 PM, Andy Dolph wrote:

> So Chris, would it be best to divide this file into 4 separate files
> with just the portion of the image that's projected on each screen and
> trust qlab to run them in sync?

Yes, that's correct, in terms of providing maximum performance. Of
course that also leads to the maximum amount of extra work to prepare
the files, so if you can get away with the in-between solution of just
assigning several versions of the cue to individual screens that might
be a good option.

Best,
Chris

sam kusnetz

unread,
Feb 5, 2010, 5:17:52 PM2/5/10
to ql...@lists.figure53.com
> On Feb 5, 2010, at 12:13 PM, Andy Dolph wrote:
>
>> So Chris, would it be best to divide this file into 4 separate files
>> with just the portion of the image that's projected on each screen and
>> trust qlab to run them in sync?
>
> Yes, that's correct, in terms of providing maximum performance. Of
> course that also leads to the maximum amount of extra work to prepare
> the files, so if you can get away with the in-between solution of just
> assigning several versions of the cue to individual screens that might
> be a good option.

i can confirm that this method works pretty well based on a recent project. just set up the cue like you were going to play it over all four screens, then copy and paste the cue three times, and then activate one screen per cue.

cheers
sam

Dave Mickey

unread,
Feb 5, 2010, 6:51:53 PM2/5/10
to ql...@lists.figure53.com
Save your files as photo-jpeg. This worked for me running 4 screens on
one mac.

The photo-jpeg makes quartz composer work less so it runs smoother and
doesn't crash the system.

Good luck.
Dave

Dave Mickey
Integrated Media Design
Sound, Video, Lighting, Interactive
714 686 1998
davem...@mac.com
DaveMickey.com
MPGinteractiveArts.com

Powered by iPhone

On Feb 5, 2010, at 1:50 PM, qlab-r...@lists.figure53.com wrote:

> Problem playing video over 4 screens

Arnoud No

unread,
Feb 6, 2010, 5:26:35 AM2/6/10
to ql...@lists.figure53.com
> Save your files as photo-jpeg. This worked for me running 4 screens on  
> one mac.

Amen to that, H264 is processor intensive; it's great for internet and
could also do for single-channel video. In all other cases use
PhotoJPEG.

arnoud

Paul Gotch

unread,
Feb 6, 2010, 12:11:14 PM2/6/10
to ql...@lists.figure53.com
On Sat, Feb 06, 2010 at 02:26:35AM -0800, Arnoud No wrote:
> Amen to that, H264 is processor intensive; it's great for internet and
> could also do for single-channel video. In all other cases use
> PhotoJPEG.

Which is disk intensive...

I'm sorry but there is no silver bullet involving video codecs, it
depends on what the disk, video and cpu hardware is, how many
concurrent video streams you are playing and what resolution you are
using.

-p
--
Paul Gotch
--------------------------------------------------------------------

Drew Dalzell

unread,
Feb 6, 2010, 3:17:49 PM2/6/10
to Discussion and support for QLab users.
It is true that it is disk intensive, but on a MacPro is a very easy
to add a multiple drive internal array. The machines that were used
for the project Dave mentioned had a single SATA drive for the system
drive and then a dual drive SATA striped array for the video files.

Drew

talkingtobrian

unread,
Feb 16, 2016, 11:20:25 PM2/16/16
to QLab
Hi Chris. Sorry to resurrect this if I'm off base, but is this still an issue in version 3? We are having sync issues with one video playing on four projectors (one projector is providing two surfaces, so a total of 5 surfaces) - so are we supposed to be running a separate file of the same video to each screen, so each file is being decoded most efficiently?

Similarly, we are hitting sync issues with audio being fired on one computer and the video from another, with an OSC command on the audio machine triggering the video playback (this is to save on mixing inputs and simplify workflow, rather than be taking audio from multiple sources).

Any insight or correction is appreciated!
-Brian

talkingtobrian

unread,
Feb 16, 2016, 11:24:20 PM2/16/16
to QLab
Oh, and we double checked - video does have audio embedded, even if it isn't going through the house, everything is at 44.1, and we are using 30fps with Photo-JPEG on video.

Andy Lang

unread,
Feb 16, 2016, 11:38:18 PM2/16/16
to ql...@googlegroups.com

On Tue, Feb 16, 2016 at 11:20 PM talkingtobrian echostatio...@gmail.com wrote:

Hi Chris. Sorry to resurrect this if I'm off base, but is this still an issue in version 3? We are having sync issues with one video playing on four projectors (one projector is providing two surfaces, so a total of 5 surfaces) - so are we supposed to be running a separate file of the same video to each screen, so each file is being decoded most efficiently?

Hi Brian,

First off, a general request to everybody, if you are replying to a four-year old thread on here, please quote enough of the message you’re replying to so that we know what you’re talking about. A lot of the members here, including all of the Figure 53 staff, read this via email, not on the Google site, so we have to go dig up the old message you’re replying to to see what you mean.

(I have recently adjusted the group settings to add a link to the thread on the Google site to all emails, which is helpful, but it’s still appreciated if you save folks that added work.)

That aside, Brian’s asking about a very old post in which Chris explained that, in QLab 2’s video engine, it was more efficient to run multiple copies of a video, one to each screen, then to run a single video to four screens.

That original discussion is here:
https://groups.google.com/d/msg/qlab/VCRsPVbljac/baa51UyYwt8J

The answer is no, QLab 3’s video engine is entirely different than QLab 2’s, in just about every way, and this advice is no longer applicable. (And, in fact, 3.1’s is very different from 3.0’s.) 

In this case, I think it’s going to be most efficient if you write to us directly, so we can gather some logs and other information from your computer.

If the computer this was happening on is connected to the internet, you can just go to the Help menu and choose “Contact support…” Then put your email address in the contact field, and remind us that it’s a follow up from your post here in the notes field, along with as much detail about the specific problems as you can give us. Finally, check the boxes that say “Send details…” and “Include logs”, and hit the button to submit the report.

If the computer is not online, you’ll have to send them manually, which is a little more work. You can get the console logs as follows:

Launch /Applications/Utilities/Console.app
Click “All Messages”
Export this as a file and attach a copy to your email.

Thanks!

-Andy


Andy Lang
@SoundGuyAndy
sup...@figure53.com

Andy Lang

unread,
Feb 16, 2016, 11:41:33 PM2/16/16
to ql...@googlegroups.com

On Tue, Feb 16, 2016 at 11:24 PM talkingtobrian echostatio...@gmail.com wrote:

Oh, and we double checked - video does have audio embedded, even if it isn't going through the house, everything is at 44.1, and we are using 30fps with Photo-JPEG on video

Sorry, I saw this after I sent my previous reply. My suggestion to contact us directly with logs still stands, but I did want to touch on this real quick. If you haven’t tried other codecs, it’s certainly worth experimenting a bit.

While QLab can play files in any format supported by AVFoundation, we recommend the following formats, listed in order of preference, for videos without transparency:

1) Prores 422 Proxy
2) Prores 422 LT
3) PhotoJPG
4) H.264 - ONLY as a last resort, and only on computers that feature hardware H.264 optimizations. On other computers, H.264 can be horribly processor intensive to decode, and is the cause of a significant number of cases of poor video performance that we see

For videos with transparency, also referred to as alpha channel support, AVFoundation only supports ProRes 4444.

Since you’re using PhotoJPG, it certainly would be worth a try using 422 Proxy, to see if that helps at all.

Also, it’s important to make sure you’re using the most up to date release of QLab. We made some important fixes to sync between multiple displays in 3.1.16, as well as some related fixes in the updates since then, so if you’re using an older release, that could be causing issues, too.

Thanks again!

talkingtobrian

unread,
Feb 17, 2016, 12:07:41 AM2/17/16
to QLab
Sorry - I saw your reply the other day, and I deliberately didn't delete the prior messages in the interest of keeping the board clean. However, from mobile and iPad I now see there is no quoting when you hit reply. :(

One question, since it will take time to go through all of this and we have a lot of brains tackling this issue - would we be best served using a wordclock to sync the two computers? My knowledge of wordclocks is a little limited, more in the recording studio. But we have some MOTU units and could, concievably, use that somehow...
-Brian

Andy Lang

unread,
Feb 17, 2016, 12:21:04 AM2/17/16
to ql...@googlegroups.com

On Wed, Feb 17, 2016 at 12:07 AM talkingtobrian echostatio...@gmail.com wrote:

Sorry - I saw your reply the other day, and I deliberately didn't delete the prior messages in the interest of keeping the board clean. However, from mobile and iPad I now see there is no quoting when you hit reply. :(

No worries…it’s all about striking a balance, just quoting enough to give context. Don’t quote an entire digest, but quote enough of the original message so folks know what you’re talking about.

And yes, some iOS clients make that easier than others.

One question, since it will take time to go through all of this and we have a lot of brains tackling this issue - would we be best served using a wordclock to sync the two computers? My knowledge of wordclocks is a little limited, more in the recording studio. But we have some MOTU units and could, concievably, use that somehow.

My turn to apologize, I’d glossed over that part of your post. There is a benefit to sending the same clock signal to both computers, and assigning the video’s audio track to the interface that’s receiving the word clock, but it only goes so far.

What it will do is keep the computers locked together so that, once those cues are started, they won’t drift from each other.

What it won’t do is guarantee that they start in sync. That’s not currently possible in QLab. The best you can do is use identically specced computers with SSDs, fast processors, as much RAM as you can, and fully optimized by following the prep guide on our site:

http://docs.figure53.com/docs/qlab/v3/general/preparing-your-mac/

That will increase the chances of both cues starting simultaneously, or as close as possible. But it won’t guarantee frame accurate sync. The only way to do that is for audio and video to come from the same computer.

Hope that helps!

talkingtobrian

unread,
Feb 17, 2016, 12:39:59 AM2/17/16
to QLab
------There is a benefit to sending the same clock signal to both computers, and assigning the video’s audio track to the interface that’s receiving the word clock, but it only goes so far.



Thanks, Andy. Luckily we are using same-spec Mac Pros, except one has more ram and different OS, one version off (Yosemite and El Capitan).

So, no chance that our hope of "this is too simple to actually fix our problem" by making four copies of the file and projecting them at the same time rather than decode the same video four times with different custom geometry" is gonna help? :)

Andy Lang

unread,
Feb 17, 2016, 1:28:22 PM2/17/16
to QLab

Just a quick note to those following along that Brian’s reached out to us directly, and we’re working with him to dig into his specific setup.

Thanks!

-Andy


Andy Lang
@SoundGuyAndy
sup...@figure53.com

--
--

Change your preferences or unsubscribe here:
http://groups.google.com/group/qlab

Follow Figure 53 on Twitter: http://twitter.com/Figure53
---
You received this message because you are subscribed to the Google Groups "QLab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qlab+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/qlab/1ea98426-9603-4599-b647-0149d07e8f1e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Reply all
Reply to author
Forward
0 new messages