matkeane

Salut Benoit - Tu me raconteras ça en détail un de ces quatre...! From your description, it sounds like the lasers are sending an x/y coordinate to Watchout, which then checks for a 'hit' with cues at that screen position and triggers the linked task if appropriate. That is great for live interaction with the screen, but I've just had another look through the Watchout manual and can't find any details about this. Is there some documentation somewhere about how to get this working? Thanks, Matthew

Thank you for the detailed response Mike - it's much appreciated!

I was hoping somebody from Dataton - maybe Mike - would chime in on this, because I would love to know more about how Compositions and Tasks work under the hood. I sometimes encounter strange behaviour with Compositions and I don't know whether this is caused by the media I'm using, the way I've structured the project, performance limitations of the machine I'm using, or some general constraint when using Compositions. I assume that Compositions add complexity to the rendering pipeline. I don't know whether Watchout pre-renders the elements in a Composition and then composites them onto other timelines (like After Effects pre-comps), or whether media cues from Compositions and Tasks are all processed and rendered in one go, but either way there must be a processing overhead. I don't know, for example, whether the dimensions specified for a Composition have any effect on the number of pixels rendered (or considered) by Watchout. Unlike say, After Effects, which crops compositions to the boundaries, if I create a 600x600px Composition but drop in a 1920x1080 media, the entire media is visible in the timeline where I place the Composition. When creating a Composition, the dimensions default to cover all the Displays (which sometimes leads to a small media item in a huge Composition) but I don't know whether unnecessarily large dimensions mean a performance hit because Watchout will process all those pixels. As to PierreLucB's questions about looping and free-run. I don't think I've tried free-run inside a composition, but I often put a Composition with a video or audio cue into a Task, with free-running and looping set, in order to be able to fade in or out of a loop at any time, and that generally works fine. One exception I've seen is with nested compositions: The other day I had a series of Compositions with title images animating into place, as I needed to reuse the same title group in several places at different sizes, and Compositions are good way to handle that. When I tried to nest the individual title Compositions into a master Composition (which was then looped in a Task for a holding-animation before the conference) the title Compositions would all play back once, but the master Composition refused to loop. So looping & free-run seems to work fine with one level of Compositions but not when they're nested - at least, not on the machine I was using. I'm not seeing a black frame at the end of a video in a composition in the way you mention. Two things: Does the Composition length match exactly the length of the video (down to a the decimals)? And secondly, perhaps the mp4 video is encoded such that the last keyframe is black - the pre-configured settings for mp4 files in compression software are often set up with web delivery in mind and are not optimised for Watchout playback. In general, I think the Free-run and Loop settings apply in Timelines, but not in Compositions - at least, they don't work for me inside a Composition, so for point 3, if you want to combine normal playback of one clip followed by looped playback of a second clip, I think you could do that in a Task, but not a Composition. And if you need to hold on the end-frame of a composition, I think you would need to add a pause at the right moment in a Task. As for pre-loading, it's true that Compositions offer neither pre-load or in-time settings, but the media still has to be pre-loaded. I assume that Watchout is scanning ahead in the timeline pre-loading media in Compositions about to be displayed, but it would be great if there was some visual indication of this. Video editing and animation software often display little coloured lines along the top of the timeline to indicate where media has been rendered and is ready for real-time playback. It would take some of the guess work out of media playback for Compositions and Tasks. As an example, a few months ago I was given a 20 minute long 1080p h264/mp4 file encoded by the client as a looping backdrop for a conference. In order to get a smooth start to the playback, I had to leave the Task on 'pause' for a couple of minutes before I hit play; any less and the playback stuttered at the beginning. I assume Watchout needed time to buffer part of the video but, other than trial and error, there was no way to tell precisely when Watchout would be able to play back that media smoothly. As another example, last month I ended up using Watchout in 'VJ' mode, basically filling the screen with video loops, overlaid with different blend modes. Although the Display PC is not the latest-and-greatest machine, at one point I had a dozen layers of HD MPEG2 looping simultaneously and (as far as I could see) smoothly. The hiccups occurred when cueing a new clip. Even with just 1 clip playing, when I triggered a second task with a video clip set to Free-run and Looping, all of the media currently playing would pause momentarily when arriving at the Pause cue in the new Task (In each Task, the video clip had an opacity tween to fade on and off, and a Pause control cue at the mid-point). So, as I said, playing 12 streams of video at the same time didn't cause a problem, but one Task hitting a Pause cue caused all the Tasks to hesitate for a moment, which I find confusing. The media is already playing, so presumably it's not a pre-load problem, but a performance limitation of the machine but, can that be the case when the machine is already playing the media smoothly? If somebody with better knowledge of Watchout internals has any hints or advice about how Watchout is loading and managing media in Tasks and Compositions, I would be interested to learn more - thanks!

I've also experienced problems before with blend modes in compositions in WO5 (layers disappearing suddenly when the composition containing them fades out) but I assumed it was because I was exceeding the playback capacity of the Display machines (either the CPU or GPU not keeping up). Is there some kind of limit to use of blend modes in Watchout itself that would cause the opacity hiccups?

Hi Walter, That sounds like a job for a script, just in case nobody has a huge number of inputs ready to copy and paste. And I agree with your feature request - a lot of tedious setup work could be avoided if there was a way to automatically create a specified number of identical inputs/compositions/tasks/etc... Matthew

Hi, I haven't had the need to change the pre-roll settings for a long time (in fact, I think it was Watchout 4 last time I did that, so things may be different in v6), but when I had a similar situation, with several large videos starting simultaneously, the solution was to stagger the pre-roll for each video, so that Watchout wasn't loading all of the media at the same time. You could try setting the pre-roll of your clips to 1, 1.5 and 2 seconds and see if that helps. If not, you may need use a faster drive for your media. You didn't specify the data rate of your clips, but make sure the total of the 3 videos is sustainable from the SSD you have. Another thought: I'm guessing that the pre-roll might need to be at least equal to the duration of the key-frame interval of your videos to ensure a smooth start, but that's just a hunch.

I'm curious about what, if any, delay you're seeing using HDbaseT systems. We were just testing some low-cost Chinese boxes at work the other day and, in my unscientific tests of waving my arms around and clapping, I could see at least a half-second delay - certainly enough that it probably wouldn't be acceptable in a live show. Do other brands perform better?

In case it could be of use to somebody, I just posted a new version of my Display Builder app on my site: http://matkeane.com/project/watchout-display-builder-softedge-calculator The new version makes it possible to add groups of displays with different resolution, rotation, etc and align them to other groups. So, if you need to align a row of HD1080 displays with another row of 720p displays, that's easy. You can also specify and calculate physical dimensions, so you can quickly check how many projectors you need to cover a certain surface area. If you need to mix displays of the same resolution but different sizes, the software will calculate the 'stage size' for you, so you can easily align a 40 inch flat screen with a 50 inch one and maintain the correct image size between the two. Once you've laid out the displays, you can export reference images with a pixel grid (useful for graphic design) or a physical grid (handy for making sure the client's logo isn't hidden behind stage props) as well as a text file ready to paste into Watchout. The software doesn't support the new display settings in Watchout 6 - Masks, Muting & Virtual Displays - but layouts of 2D displays for Watchout 5 seem to be working fine in WO6.

I had similar experiences in a fixed installation sending TCP cues to a Crestron system a few years ago. The Watchout 4 setup used a Producer machine, rather than a Display cluster and we occasionally saw problems with missing cues. After some testing with Wireshark scanning for the commands sent to the Crestron we found two things: Firstly, as Alex Ramos says, if two cues were sent at the exact same time from a timeline, sometimes only one would get triggered, so we staggered all commands by 150ms and that sorted that problem. Secondly, after testing the show several times, it turned out that the Producer PC would sometimes skip cues in the timeline if it was under heavy load. This was a project with around 30 HD outputs and, when the Producer machine was set to 'Best' preview quality, cues would get skipped. It's not that the Crestron was ignoring cues; they never showed up in the Wireshark logs. With preview quality set to 'wireframe', or 'video as thumbnails', all cues were sent successfully. For some commands, like cowboyclint suggested, we doubled up cues with around 250ms delay to be sure they got through. There was at least 10 minutes between the start/end of each show when commands were being fired, so I assume Watchout disconnected in between. Once a show ended though, a series of commands would be sent for lighting controls, door commands and machinery within a minute or so, but these were susceptible to the same problems.

Vollmers: Unfortunately there isn't an easing parameter for setInput (unless Dataton have something up their sleeves for an update). That's why I'm using an external app to calculate the acceleration/deceleration and send a series of updates to Watchout. The difference in v6 is apparently that it copes better with frequent updates - it's quite easy to swamp v5 with network messages, which causes playback to stutter.

I will do some experiments with the new version of Watchout sometime when I can borrow a friend's v6 dongles. I was just surprised to see such a big difference in performance between production mode and a display cluster - animation of 10 layers was smooth as butter when talking to the production PC, but the Display machine started having trouble with only 2 layers. One last question - Are there any plans to add external control of compositions? I was trying to move two cues (an image and a drop shadow behind it) together by linking them both to the same inputs, but there was a little bit of wiggle between the two - maybe the position being rounded to whole pixels?

After reading this other discussion about real-time control of Watchout (http://forum.dataton.com/topic/1843-realtime-tracking-in-watchout/) I continued my experiments. The first thing I did was to switch to sending commands via UDP, rather than TCP, which should create less network overhead (and seems to be working successfully for the Spikemark software). So, I have 10 layers moving around in real-time (updating x,y,z and opacity information) and everything is nice and fluid. At least, it is as long as I run in Production mode and send the commands to the Producer machine. If I switch to Cluster mode, and send commands directly to the Display machine, everything starts stuttering and lagging again. So the very same player copes fine when controlled by a Producer, but staggers to a crawl when running as a standalone Display. Is this expected behaviour? Does using a Producer mean that the producer PC is dealing with the network overhead so that the Display machines don't have to?

Normal, no, but common, yes. There's another discussion about this issue in Watchout 5 (http://forum.dataton.com/topic/1518-jump-to-run-in-aux-timeline/). Putting your looping content in a composition, and then putting that comp into an aux timeline for fade up/pause/fade out, seems to avoid the problem.

Hi Mike, I'm using Watchout 5 at the moment, but it's good to know that I would be able to pass in an array of values to Watchout 6. I'm curious about the difference in playback when animating cues on the timeline and using setInput - I can animate 20 cues simultaneously in the timeline and playback is fine; if I try and do the same thing using inputs, things start to lag and stutter pretty quickly. I was experimenting with using a force-directed graph to position text in a word cloud, with new layers pushing other layers out of the way to make room, before the 'cloud' settles down into a new layout. I guess I could use my external app to calculate the new layout positions and send a single command for each layer, but it would be nice to get a smooth moving with easing, which is why I was trying to send a stream of position data. Assuming I stick with v5 for the moment, would Midi be a better choice for low-latency position updates? Thanks,

Hi, I've been experimenting with external control of Watchout cues using setInput. Setting the tween time with the third parameter works fine, except that the movement is linear, with no easing, which looks a bit harsh. I don't suppose there's a hidden 4th parameter that enables ease-in-out? I have been playing with sending position data from an external app to try and achieve a smoother move with easing. First by trying to update the generic input used for position at 60fps. This works pretty well with one layer but, as predicted by Mike Fahl in this post - http://forum.dataton.com/topic/1262-interactive-inputs-are-freezing-up-watchout/ -once you start sending updates to a few more layers, things slow down and eventually crash. My next approach was to split the move into segments, and send setInput commands every few frames with a tween time, to try and approximate an easing curve while sending fewer network commands. This works a bit better, in that things don't lag as badly, but the change in speed between segments is noticeable. Even using this approach, and staggering the network commands, once I had 10 layers moving around, things got pretty slow again. So a couple of questions: Is the slow-down when moving multiple layers due to overhead in the network protocol? If I use a midi controller and link the sliders and knobs to Watchout cues, I don't notice the same slowdown. Rather than using a generic input, would I be able to achieve smooth changes without bogging down Watchout by sending Midi updates through a virtual midi interface? Thanks for any ideas...

I have had text files for the Korg NanoKontrol 1 and an Akai LPD8 hanging around on a USB key for a while. This thread finally prompted me to put them up on Github where anybody can grab them (and I can find them again when I need them!) - https://github.com/matkeane/watchout-snippets

Hi, Just wandering if anybody has any experience using the Matrox VS4 Quad SDI card as a live input with Watchout. http://www.matrox.com/video/en/products/vs4/ A client has two of these cards in PCs used for multi-track recording and streaming of live events with Wirecast. I was able to do some tests with Watchout 5 on these machines and the card shows up as a live input source - actually four times, one for each input. In my quick tests with 2x 1080i50 SDI video inputs and a 3rd live source from a PC, I saw 2 frames of delay between the source and the Watchout display (checking on a video recording with source and output visible) which doesn't seem bad compared to alternatives from Blackmagic. The cards seem reliable for long (2-3 hour) multi-track recordings - I just wondered if anybody had experience using them as a live source with Watchout?

I've uploaded a simple WO6 test file and media to Dropbox: https://www.dropbox.com/sh/0h5aj5hxnk4pt6e/AACu1AppumF6T3dEW4NwHWV1a?dl=0 The project contains 3 video cues on separate layers: The first has no special settings; The second is set to free-run/looping; The third uses the new playback speed setting in WO6. There are two text files - 'media video 01 wo6 project.txt' is the unzipped DFC_Data file (renamed for clarity); 'media video 01 wo6 main-timeline.txt' is the result of pasting the content of the main timeline into a text editor. The main project file structure contains project preferences, displays and the media window content, which you may not necessarily need if you just want to get the timecodes and filenames of cues in the main timeline. Hope that helps. By the way, at home I usually program shows using Watchout in a virtual machine on my Mac - works fine. @Mike: Thanks to your instructions in the other thread, I now have Watchout 5 and 6 running on the same machine, and the comments in the WO6 project data are very helpful. Thank you!

Hi Mike, I'm downloading the WO6 Beta now, so I haven't had a chance to install and play with it yet, but extra comments in the data sounds great - I've sometimes got lost while wading through deeply nested properties, so that will be really helpful! doctorhandshake: I tend to copy/paste data in and out of WO, rather than attempt to modify the show file directly, so there's always the option to undo. Also, if you mess up the syntax while pasting, WO seems to ignore anything it doesn't recognise, so malformed data just silently fails, rather than corrupting the show. An easy way to get started is to drop a few cues on the timeline, then select everything and paste into a text editor.

Hi, There's no native 'import/export as XML' function (maybe in WO6?) but it is possible to get cue data in and out of Watchout using text files and some scripting. I toyed with this idea for a project where the video content was being rough-edited in Final Cut before being added to the Watchout show, but never got around to it. Support for XML import/export could simplify some of the repetitive parts of show creation - Maybe you should add a post to the feature request thread: http://forum.dataton.com/topic/794-feature-requests-post-here/

Thanks for the mention, Jonas! The primary function of Display Builder is to lay out Displays for Watchout, but I added simple colour & greyscale bar generation as I find them handy for adjusting the softedge curve. I wasn't even sure whether to keep that function in the next version, but if people find it useful... For more complete test charts, I like the Belle Nuit video charts available for free download here: http://www.belle-nuit.com/test-chart The downside being that they only exist in standard video production resolutions.

My experience with the W7000 was on a client's machine, and I don't have access to it at the moment, so this is from memory. The card was supplied, and I think also installed, by Vidémus - the Dataton partner here in France. They supplied a PDF fie with screenshots of how to configure the EDID (which I would upload if I could see how). In the ATI FirePro control panel, under the 'AMD FirePro' tab and 'EDID Emulation' options: the 'Force EDID Emulation' checkbox is checked. Then, after expanding the 'Display Connections' header, I've selected the outputs I want to configure and hit the 'Manage EDID' button at the bottom. From there what I have generally done is either load an existing EDID configuration from a file or, if I'm setting up a new configuration, I've first saved the EDID file from the first output, and then applied the EDID file to the other outputs. Although I've used the machine on several events now, I've actually either been using 1080p60 outputs or 1920x1200p60 outputs, so I've been swapping between the two EDID files. Last week, while preparing a show on the client's premises, one of the LCDs was intermittently losing the input, so I spent a few minutes swapping cables and DisplayPort adapters while Watchout was playing the show and it didn't budge (turned out to be the DVI-HDMI adapter plugged into the screen). The only drawback of the card I've seen so far is that it's not necessarily easy to force some resolutions using just the card. I spent an interesting 15 minutes at one event trying to set the outputs to 1080p50 while connected to a Christie projector which kept forcing the card back to 60Hz. When I have some time, I think I will borrow an external EDID emulator and save a bunch of configuration files that I can reuse with the card.

What kind of media file are you trying to import? I've experienced some problems with some PNG files (titles created in Illustrator) crashing Watchout in the way you show but, generally, recreating the media file sorted things out. I never managed to pin down the cause but it seemed to happen most of the time with PNG files with simple content and transparency.

The ATI cards seem pretty stable once the EDID is locked in. The ATI configuration panel allows you to 'Force EDID emulation' on the chosen outputs and, once that's in place, I've unplugged and swapped cables without Watchout noticing a thing - the show just keeps playing. Without the forced EDID though, it just behaves like a normal graphics card and tries to detect the output format, which might explain the instability others have seen.

@JFK: Thank you for all that information - it's very helpful.