Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
Displays

High End Video Capture? 75

Posted by Cliff from the hardware-image-acquisition dept.
A reader asks: "I work for a very well known company specializing in Game Engine Middleware. Recently we've been trying to gather together marketing material for some new products, and one step towards that end is capturing high resolution gameplay footage (1280x1024) into some kind of movie file for editing. According to the 'experts', the best solution is to scan convert the DVI out into HDTV 1080p, and then HD capture it back into another PC for editing. Surely all this conversion to 'broadcast' quality is pointless - has anyone come across a pure DVI capture solution?"
This discussion has been archived. No new comments can be posted.

High End Video Capture? More

High End Video Capture?

Comments Filter:

  • Software, but... (Score:3, Interesting)

    by TheSHAD0W ( 258774 ) on Friday April 07, 2006 @12:57AM (#15082124) Homepage
    You'd probably be better off listening to the experts, because the bandwidth off a high-res video connection is monstrous, but...

    I'm thinking it might be possible to intercept the stream of data the game is sending to DirectX or whichever 3D library you're using, and record it. This data stream should be orders of magnitude lower than the actual video data, and you ought to be able to record it without much disruption to the game performance. Once you had the data, you could then re-render the game play frame-by-frame, and then convert it to video and compress it.

    Hum. You might be able to hack this into your game code; but if you can do it externally, it might be a saleable product.
    • That's a good idea; but here's a question... if you re-rendered it using something like Renderman (and it ended up looking a lot better than the actual game) would that be considered false advertising? I guess you'd have to re-render it frame-by-frame on a similar card, and individually capture each frame. It wouldn't be realtime, but you could play it back at realtime.

      But also, aren't 3D GPU's state machines? i.e. just because you have all the vertex data and everything doesn't necesarily mean you know

    • Re: (Score:2, Informative)

      by account_deleted ( 4530225 )
      Comment removed based on user account deletion
      • Comment removed based on user account deletion
      • Fraps doesn't cut.

        Imagine you are playing a computer game on your PC that is designed to use 100% of all resources available on the computer (CPU,GPU,RAM) do you think there is enough resource left to capture/compress/save images to disk ?

        Who have tried that know this, the quality is miserable.

        For GUI applications, there are tons of screencapture tools, which work fine since frame rate is not important. But for those who need a high quality video capture, frame drops are not acceptable.
      • It is because the spcs are what is required.

        This is a typical /. response:

        You could use X. It doesn't meet your requirements.

        I know I oversimplify what you stated, but the bottom line is: He is talking business here. Rule #1 will be that the solution will have to meet the requirements.
        • I know I oversimplify what you stated, but the bottom line is: He is talking business here. Rule #1 will be that the solution will have to meet the requirements.

          There probably is no solution which exactly matches his requirements in this case. Bearing that in mind, which of the following answers do you think is most helpful?

          A) Tough shit, there's nothing that meets your exact requirements so you're screwed.

          B) Have you considered Product X? It doesn't quite meet your requirements, but with a bit of work on y
          • What do you mean, moving to? I used to manage a marketing department (now I do Business Development, so no management in the hierarchical sense anymore)

            However, things have many shades of grey to me :-). I do not think the given answer was wrong completely, but it is taking to many steps at a time in my opinion. I would favor:

            1 There is no solution that meets your requirements (If that is the case, other comments seem to suggest there is)
            2 Would it be acceptable for you to loosen some requirements?
            3 If so,
        • Requirements are often mis-stated. In this case they aren't even explicitly stated: The problem is ill-posed. It is always worthwhile, when a nearby solution is cheap, while a solution that mindlessly adheres to the specific expression of the requirements is costly -- perhaps prohibitively costly -- to ask whether the cheaper solution would be satisfactory. The question cannot be asked unless the cheaper solution can get on the radar.
    • Easier method would be to just grab the frames as the game renders them. Hook into the D3D/OpenGL calls and get the finished frames and save them. If you also hook the timer functions the game uses you can make it run at whatever framerate you want without having to worry about how long it takes for you to save those frames.
    • back in the 90s games such as Doom/quake made it so you could record your gameplay and play it back exactly. instead of storing the video data it stored the gameplay data. a bit like making a savegame for every frame. the id games actually did this by simply recording every event as it occurs (player steps forward, monster 1 turns left 10 degrees, rocket 12 hits wall). the playback was made reliable by a random-number generator who's output was always the same from initialisation... the "randomness" of the
      • I believe the idea behind thouse recording formats were to capture player input on the grounds that all other variables need not be recorded because they will synce if the input synces. Less like a savegame, more like an input macro.
  • 20 minutes with Google and all my fantastic video expertise (!) and I can't find a thing for you. All the HD stuff uses HD Serial Digital Interfaces, not DVI (except for monitoring). Those HD boys work in a different world to us.

    I looked at BlackMagic, AJA and Canopus.

    Capturing HD takes a lot of grunt and space.

    I'll be interested to see the final answers here.

  • Lossless capture solution (Score:5, Informative)

    by Samir Gupta ( 623651 ) on Friday April 07, 2006 @01:18AM (#15082269) Homepage
    Step 1: DVI (Analogue or Digital)->HD-SDI - XDVI-20s

    http://www.doremilabs.com/products/XDVI-20.htm [doremilabs.com]
    http://www.onevideo.co.uk/xdvi20s-p-359.html [onevideo.co.uk]
    (In the UK £2,687.23 inc VAT)

    Step 2: HD-SDI capture board - Blackmagic decklink HD pro 4:4:4

    http://www.blackmagic-design.com/products/hd/ [blackmagic-design.com]
    http://www.onevideo.co.uk/decklink-hd-pro-444-p-11 5.html [onevideo.co.uk]
    (In the UK £959.98 inc VAT)

    There are many other alternatives to this. This is just one suggestion that I have tested to work.

    For my capture PC:

    Opteron 254 (2.8ghz)
    Tyan Thunder K8WE
    Adaptec PCI-X Ultra 320 SCSI Raid controller (39320 series)
    4 x 300GB 10,000rpm Seagate SCSI disks running as raid0 (6-8 would be best)
    New Nvidia graphics card
    2GB ECC RAM

    • for step 1, you can also use a miranda DVI-Ramp
      http://miranda.com/product.php?i=235&l=1 [miranda.com]

    • Re:Lossless capture solution (Score:4, Informative)

      by ickypoo ( 568859 ) on Friday April 07, 2006 @03:48AM (#15082665)
      You really don't even need that many format conversions. Frame capture cards exist for a reason, and there are DVI-native capturing solutions available such as the Unigraf UFG-03 [unigraf.fi] and the Foresight Imaging AccuStream 170 [foresightimaging.com].

      Relatedly, there's actually quite a market for VGA-level capture devices. Anystream and Sonic Foundry both market products that will capture video and VGA, and combine them into various "rich media" presentations. At work we use Anystream's Apreso system to combine video of professors with their live powerpoint doodlings, and present it as archived online lectures. I fully expect that as DVI becomes more common, DVI-capturing solutions will likewise become more common -- if for no other reason than to tap into the same market that exists for VGA capturing.
      • I built a Suttle XPC with a Foresight AccuStream board and 3 hard drives, two of which are in RAID0. The included Intel ICH6 chipset was perfect, because RAID is built in and the SATA drives don't use any PCI bandwidth (you'll need it for the capture card!) The AccuStream captures both analog and digital at up to 1600x1200. (However for me 1600x1200 digital doesn't work - it gets noise from signal loss) The whole box was about $5k, maybe a little more, and most of that is the capture board. This was ab
    • Or just stick it in a G5 with a PCIe black magic card and drop it on an xserve raid with 14 250 gig drives over fiber channel... A little pricey, but it looks nice. especially if you put the raid in an xtrovert vertical rack.

      http://www.xrackpro.com/xtrovert1.htm [xrackpro.com]
  • I know I'm not adding a solution here, but I was attempting to think outside the box here:

    PVR's? just off the top of my head, can they record from aux sources? most have DVI output, but do any have DVI input? that would be a nice pure DVI solution if such a device existed. or DVI-To-HDMI.... keeping it digital.

  • Why do you need a hardware solution? (Score:5, Informative)

    by Guspaz ( 556486 ) on Friday April 07, 2006 @01:26AM (#15082294)
    Have you looked into FRAPS (http://www.fraps.com/ [fraps.com])? It doesn't quite meet your resolution requirements, but still gets you most of the way there.

    It can record at 1152x864 (4:3) or 1280x720 (16:9) as a max resolution.

    1280x1024 is only about a third higher resolution. Perhaps there is some technical limit that prevents fraps from passing one megapixel per frame (both supported max res are slightly below that mark), and 1280x1024 is 1.3 megapixels. But maybe they just picked a megapixel as an arbitrary ceiling to prevent customer complaints from slow performance.

    I don't know anything about the internals of FRAPS, but it seems ideally suited to a dualcore system.

    I suggest you contact the FRAPS people and ask them:

    1) If a special build can be produced that supports 1280x1024
    2) If FRAPS can take advantage of a second core (Game on one, FRAPS on the other) for such intensive recording

    The demo videos are impressive. The UT2003 one at 1024x768 is just the intro and title screen, but the 800x600 Doom 3 demo is a minute of gameplay, and it doesn't seem to be dropping any frames.
    • Have you looked into FRAPS (http://www.fraps.com/ [fraps.com])? It doesn't quite meet your resolution requirements, but still gets you most of the way there.

      It can record at 1152x864 (4:3) or 1280x720 (16:9) as a max resolution.

      Is there anything comparable to this in the Linux world? This looks like Windows-only software.

      I'd love to be able to record some tutorial videos for software I've been writing, but some of it is high-motion video and/or OpenGL. The closest I've been able to come so far is to run everythin

    • I was going to post this right when the topic hit, but the database was down :(

      "It doesn't quite meet your resolution requirements, but still gets you most of the way there. It can record at 1152x864 (4:3) or 1280x720 (16:9) as a max resolution.

      suggest you contact the FRAPS people and ask them: 1) If a special build can be produced that supports 1280x1024 2) If FRAPS can take advantage of a second core (Game on one, FRAPS on the other) for such intensive recording"

      Fraps can record at a much higher

      • Fraps can record at a much higher resolution than your 1280x1024 stated, as of version 2.7.0 (6th Nov 2005) it can capture video up too 1920x540 and 2160x480 resolutions. One think that can help with your video capture is a faster hard drive or setting up a RAID 0 array to record to.

        Those are not higher resolution. Notice that as the horizontal resolution increases, the vertical resolution decreases.

        1152 * 864 = 995,328
        1280 * 720 = 921,600
        1920 * 540 = 1,036,800
        2160 * 480 = 1,036,800

        Do you see the trend here

  • AccuStream 170 (Score:3, Informative)

    by Wesley Felter ( 138342 ) <wesley@felter.org> on Friday April 07, 2006 @01:42AM (#15082349) Homepage
    I can't vouch for it, but the $3,000 AccuStream 170 [foresightimaging.com] is the only DVI capture card I've seen. It looks like the cost is almost a tie between DVI capture and DVI -> HD-SDI capture.
    • wrong...
      ncast has an offering that's 1/2 that price..
      http://ncast.com/telepresenterDigi.html [ncast.com]
      • $1500:
        Up to 30 fps. XGA and 25 fps. SXGA

        $3000:
        For example, a video signal of 1600 x 1200 x 60 frames per second is 162 MHz. AccuStream 170 achieves a maximum video streaming rate of 30 frames per second from this video format.

        Hrm. Maybe there's a reason one is $3k.
    • I'm quite familiar with the Accustream 170 - my company specializes in webcasting and this is the card we use, built into a Shuttle XPC, at conferences to capture Powerpoint presentations. It's pricey, but image quality is very good. That being said, there are drawbacks: 1) The card's software in its current release isn't so good at automatic format detection. You need to run it through a calibration routine every time you feed it a new signal/resolution. Switching sync during capture throws it off if

  • the solution is readily available (Score:3, Funny)

    by Trouvist ( 958280 ) on Friday April 07, 2006 @01:54AM (#15082393)
    There is a way to do this where you don't have to write a single line of code, it is full resolution (up to 1600x1200) and can be done on the same machine you are creating the stuff on through hardware (not recommended because of bandwidth issues). I have done it. My NDA says I can't say how. Sorry.
  • http://techpubs.sgi.com/library/manuals/4000/007-4 663-002/pdf/007-4663-002.pdf [sgi.com]

    "
    The Silicon Graphics SGC option is a video frame capture PCI-X card that allows a stream
    of digital computer resolution video (as compared to Standard Definition or High
    Definition Digital TV signals) to be read into the memory of a Silicon Graphics Prism
    system. The computer resolution video is delivered in a digital form to the card via a
    single link DVI-D connector.
    The card itself is seen as an OpenML 1.1 device, and can be progr
  • Unless you're doing this a lot, you probably can find a facility that will have the appropriate scan converters. The easiest way is probably going to be DVI->HD-SDI->Capture Card. There are lots of places that have this (expensive) equipment. You're going to need very fast disks - uncompressed HD is huge.
  • I know this might not be entirely useful, but we've got a card from SGI that does this:

    http://www.sgi.com/products/visualization/media/di g_media.html#graphics_capture [sgi.com]

  • IIRC there are a few manufacturers who have firewire attached standalone harddrives. Plug DV right into them and big-fiddle, you have files.

    Maybe peruse Markertek's website http://www.markertek.com/ [markertek.com]

  • Try the DGy system (Score:1, Informative)

    by Anonymous Coward
    We recently had a similar situation. We ended up going with an all-software solution, but screen-scraping may not be for you. One hardware solution we looked at, however, was the DGy system made by RGB Spectrum. They claim 30fps at 1280x1024 resolution, plus these boxes will do sound as well, and you can just FTP the recordings off them afterwards. We ended up not going for them because they were overkill for our needs. I have no idea on pricing, but I'm guessing that's not your primary concern.
  • Take a look at http://www.unigraf.fi/?page=64 [unigraf.fi] We've been using an earlier version to capture analog video at 1280x1024@67Hz. It does support DVI input; we just haven't needed to use it. It was the only board we found that supported Linux, and it has worked very well for us. A bit expensive, but it is a very nice board.

  • start at the end (Score:2, Informative)

    by bigmo ( 181402 )
    You really need to decide exactly what you want to end up with before you start talking about how to start. The point of all this is to have something to show people.

    Will you stream it from the web? -- I doubt you'll be able to stream high def very well
    Will you distribute DVDs? -- There's not much use for HD here for a while
    Will you show it at trade shows? -- Renting an HD deck & plasma will be extremely pricey

    For a real world solution, go to an Audio Visual
  • What if you used a multi-core system (probably an AMDX2) and streamed the output via gigabit LAN?

    Since you're not writing to HD you would just need a bunch of RAM to store frames while it gets streamed. Then you can use another system in a striped RAID-0 configuration to store the output quickly.

    My other vote would go towards a frame-by-frame rendering situation. 1280x1024 is small potatoes for current graphics hardware. The resulting video won't be something you can't get in realtime.
  • Why 1080p? Is the display going on a 1080p TV?
  • Just show your footage on a huge LCD and get a room full of people to stand in front of a guy with a shaky camcorder. Then watch as every games newssite discusses your video, which is near-unwatchable.
  • This is a job for those idiot coders in the geek pen 40 feet away from you, you know.. those funny looking guys with posters of scantily-clad night elves on their cube walls.

    Seriously, you've got "Game Engine Middleware" which is just a fancy way of saying "graphics and sound engine". Is it that difficult to just tap the output of your own in-house graphics renderer and send it to a file ? Then all you need to do is encode it to something slimmer like h.264 or WMV HD.

    If you're too lazy (or your geeks are
  • Look at Matrox, specifically at Matrox Axio product line on MAtrox Video page. Still strong in this market...

Slashdot Top Deals

Pascal is not a high-level language. -- Steven Feiner

Close