Personal View site logo
Make sure to join PV on Telegram or Facebook! Perfect to keep up with community on your smartphone.
Please, support PV!
It allows to keep PV going, with more focus towards AI, but keeping be one of the few truly independent places.
The CGI and VFX thread!
  • 106 Replies sorted by
  • hishimaruhishimaru

    Nice thread! I was about to post the Final Destination clip, but looks like @Vitaly beat me to it. It was a gruelling yet intensely fun time working on that film. I'll see if I can get the breakdown reel we made at the studio and post it up sometime soon. @bwhitz Nice ship design. Are you using an HDR for any image based lighting? It's the quickest way to integrate CG to a plate matching the lighting. Shooting spherical HDRs on set can be rather tedious though

  • nomadnomad

    Since you don't need too much resolution, you can use a 180 degree fisheye on a FF camera and do the HDR for image based lighting with only two or three exposures.

  • johnnymjohnnym

    For those interested in 3D tracking, there's a 24 hour sale at http://www.thefoundry.co.uk/ for Cameratracker for AE.

  • bwhitzbwhitz

    @hishimaru "Are you using an HDR for any image based lighting?"

    Nope. I'm planning on doing that for my next composting test though...

    "It was a gruelling yet intensely fun time working on that film."

    Sounds cool!

    @pinger007 "I'm trying to perform 3d camera tracking with GH2 footage in Syntheyes. For the back plate settings, should I use the full GH2 sensor size of 17.3x13.0mm, or should it be something different since we're shooting 1920x1080?"

    Should just be whatever the FOV is equal to. I actually used my 7D for the shot I posted... and it was shot with a 30mm lens. The C4d camera is FF35mm back by default, so I just set it at about 50mm for the "lens". I could have changed the back to 25mm, and the lens to 30mm... but I don't think it matters as long as the FOV is equal at the end.

    Just found a cool light wrapping tutorial... !

    http://vfxhaiku.com/2009/10/how-to-create-light-wrap-and-edge-blur/

  • pinger007pinger007

    Shouldn't trust the software to figure out the back plate since mistakes are common and can ruin a track.

    Here's what I've figured out (please chime in with corrections):

    17.3x13mm is gh2 sensor size for taking pictures

    in video mode it will be HD, & HD is standardized to be 16-9 in aspect ratio, therefore camera sensor will crop horizontally to maintain 16-9 aspect ratio

    So gh2 video size will be 17.3 x 9.73125

    (17.3*9)/16 = 9.73125

    As for ETC mode... I am not sure about my calculation but I think the sensor size might go to

    6.67524115755627 x 3.7548231511254

    17.3*1920/4976 = 6.67524115755627

    Does all this sound correct?

    Thanks.

  • bwhitzbwhitz

    "Shouldn't trust the software to figure out the back plate since mistakes are common and can ruin a track."

    You don't... you just make sure it's on the correct FOV. Actually I don't think Camera Tracker required a back-plate setting... just C4d.

    What software packages are you using BTW? Yours may have a different setup...?

  • pinger007pinger007

    I utilize Syntheyes and PFTrack to do my 3d tracking, and then bounce it back to Maya and integrate that with either Nuke or AE.

  • thepalaliasthepalalias

    Here are a couple simple things that might be helpful to beginners in regards to getting a nice composite. They are really basic today but back when I cut my teeth, they got ignored frequently on several high profile projects.

    • Look carefully at the saturation and luminance levels of the "live" elements when you create the computer ones. Even if you can't use advanced techniques like image-based lighting/global illuminations/mirrorball HDRs/etc. you can do a lot by just eyeballing things. A good rule of thumb: when in doubt, make sure your generated elements do not exceed the saturation of your live elements. It draws more undesirable attention to the gap between them if they are over-saturated than under-saturated.

    • Always match framerates between different compositing elements. If you look at Saturday morning cartoons from the early 90s, you can immediately tell when they started doing "effects in post" because the framerate was higher than for the normal animation.

  • bwhitzbwhitz

    @pinger007

    Ok, yea. The setup must be a little different then. But as far as the math, I think you've got it... except the imaging area of the GH2 in video mode is actually 19 mm x 10.7mm. The sensor is actually wider in video mode than in stills... check out this diagram:

    http://forums.dpreview.com/forums/read.asp?forum=1041&message=36396593

    ...it's an interesting sensor.

  • pinger007pinger007

    @ Vitaliy_Kiselev

    Can you confirm the above statement? I've never read it before (didn't know the sensor can ”grow” larger for video mode). Are the forums at dpreview usually reliable? Thanks.

    Thanks for the link bwhitz. Hope to learn more on this subject.

  • tedted

    I concur regarding using 2D motion blur like ReelSmart. We wrote our own at VisionArt, as 2D motion blur was much faster than rendering it 3D. We did a bit of a hybrid approach, where we used 3D data from Houdini to help calculate the 2D blur for increased accuracy, particulary if objects were moving toward or away from camera. ReelSmart does a nice job as an off-the-shelf solution.

  • thepalaliasthepalalias

    @pinger007 Panasonic has been using "smart" multi-aspect sensors in many of their cameras for years. Even my old TZ5 was similar. The 16X9 mode uses the widest section of the sensor and the output dimensions (in terms of pixels) reflects that while the 4X3 mode has the highest overall resolution.

    Note that the diagram refers to the GH1, not the GH2, but the concept is exactly the same.

    Yes, the 16X9 mode (whether video OR stills) goes wider than the it does in 3X2 or 4X3. So no, it does NOT grow for video. If you use the 16X9 mode in stills, it uses the same section of the sensor.

  • thepalaliasthepalalias

    @bwhitz I looked at that diagram. It doesn't say anything about video mode being wider than stills - just any mode that uses a 16X9 aspect ratio, whether video or stills. :)

  • pinger007pinger007

    @thepalalias

    I understand multi-aspect sensors. The confusion was in the statement regarding video mode being wider than photo mode.

    So assuming 19mm x 10.7mm (21.81 diagonal)... The question is, what size is the part of the sensor used in ETC mode? I calculated it to be about 7.31 x 4.12 (8.39 diagonal)

    Is the sensor 19mm or 18.89mm???

  • thepalaliasthepalalias

    @pinger007 So far the only person in this thread that has said 19mm [EDIT: clarified they meant rounding to the nearest mm]. But you can quickly verify for yourself whether the video mode has a different FOV or not.

    Put the camera on a tripod, shoot a frame at 24H and then (without moving the camera) shoot the same frame as a 1920X1080 still (that would be the "S" setting in 16x9 mode). Then you can compare the FOV of the two shots.

  • thepalaliasthepalalias

    @pinger007 If the diagram at http://www.bmupix.com/journal/2010/9/21/gh2-my-next-camera.html is accurate (which was the GH2 specific link mentioned in the one @bwhitz specified), then it looks like it should be closer to 7.25mmX4.09mm, not 7.31mmX4.12mm.

  • bwhitzbwhitz

    "So far the only person in this thread that has said 19mm also linked to a diagram that didn't coincide with the statement."

    Ok... 18.8. No need to be so literal. I was rounding up.

  • thepalaliasthepalalias

    @bwhitz

    You're right and I wasn't trying to be anal. The only reason I got specific was because pinger007 was trying to calculate based off of that rounding and appeared to be getting either confused or frustrated (though I may have misread that). I was just trying to say that the only number suggested so far was the one on the diagram you (helpfully) linked to. :)

    I've edited my post to better reflect that.

  • AlexAlex

    here is some stuff from last year where we used GH2 footage for 3D tracking. Worked well:

  • pinger007pinger007

    Neither frustrated or confused (please don't assume), just wanting specifics. DPreview specifies an 18.89mm sensor, whereas Panasonic specifies a 19mm. Just wondering which one is correct.

  • thepalaliasthepalalias

    @Pinger007 "Is the sensor 19mm or 18.89mm???"

    In reference to the use of three question marks instead of one I said:

    "...appeared to be getting either confused or frustrated (though I may have misread that)."

    Then you responded: "Neither frustrated or confused (please don't assume), just wanting specifics."

    Obviously I misread you, but no assumption was made. :) Best of luck getting confirmation about the specification.

    Where does Panasonic indicate 19mm? It might be helpful in sorting this out to see whether they specify things to the nearest tenth of a mm elsewhere in the document (and if they only go the nearest mm, I would simply use the more specific value).

  • thepalaliasthepalalias

    @Pinger007 You probably already encountered this but just in case, I thought I'd share. One guy said "Panasonic hasn't, as far as I know, released the exact true width of the GH1/GH2 sensor, but from pixel calculations we think it's using an area of 18.9 x 10.6 or thereabouts."

    If you found official specifications from Panasonic, they could be really useful not only on here, but on other forums like that. So I'm wishing you the best of luck with all this.

  • bwhitzbwhitz

    Well, you could always use trial-and-error then... Try a few test tracks and see what they look like. I doubt that a margin of .1-.2 is really going to effect the outcome that much.

  • Ralph_BRalph_B

    @bwhitz Nice work with the Skycar. Question: Just before it takes off at high speed, it does a little anticipation, where it moves back a bit. Did you do that by hand or did Cinema 4D do it automatically?

  • bwhitzbwhitz

    @Ralph_B "Question: Just before it takes off at high speed, it does a little anticipation, where it moves back a bit. Did you do that by hand or did Cinema 4D do it automatically?"

    That was done by me intentionally... it just seemed more natural.

Add a Comment