Create a movie in 3D (meaning stereoscopic)
Birdie2
Posts: 24
Hi there,
I have a short Trailer/Animation created and would like to render it in a way that I could play it on my home Theater System in 3D.
How would I have to start this?
Basically I think just renidering the movie twice with a second camara Close to the regular one. but how to bring this in a 3D-playable file.
Suggestions?
Reagdards
Stefan
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Yes, render twice, one for left, one for right eye and merge with StereoMovie Maker.
StereoMovie Maker: http://stereo.jpn.org/eng/stvmkr/index.html
Sounds easy.
How far appart from each other should these two cams be?
Can I somehow make the second cam "automatically follow the existing one"?
Otherwise will probably very hard to have them move synchronized, right?
What Software are you using? Studio? Carrara? Poser?
In Poser you can take a simple prop, make it invisible, create two cameras, position them, parent them to the prop. Now animate the prop and record with both cams.
Daz studio you can use a null object
Carrara can group
I am using DAz (4.5).
Since the Trasiler is (as 2D Video) already finished, Wouldbe best, if possible, to simply somehow link the second cam to the existing one.
What do you mean with "null" object?
hmm you might try creating a visible object and set it just beside the first camera. Now make the camera the parent of the object. now set up camera 2 beside the object and make the object its parent. that might work but the best way I can think of is to make the object the controlling parent to both cameras.
OK, solved it like this now:
1. I copied all keyframes of the existing cam
2. created a second cam
3. made a Group out of the two cams
4. pasted the keyframes to the grouzp
5. removed the keyframes from cam1
So I now have two cams moving in parallel.
One more question:
for the best 3D effect, should the cameras move and "look" really parallel or shouldn't they both look at the same point.
hard to explain: At the moment they are focusing on iternety, but shouldn't they look at one focus point where my objects are?
Does this even make a difference in the result?
Seems the StereoMovie Maker has some issues with the huge files.
I am working in Full HD 1920/1080, so just 350frames result in about 2GB auf data.
Any other Software which should be able to do that?
I was usually working with AvidStudio to put separat files into one and to add Audio tracks.
Unfortunately my Version does not Support 3D Video.
Does anyone know if the current version (Pinnacle Studio 17.5) could build a 3D SbS out of two separat files?
stereomoviemaker can use image series if I recall
I have used it to do quite large stereoscopic movies,
I always render image series
any uncompressed avi is quite large anyway, my average working files are usually 5GB or more for 30secs totaling sometimes 100GB for a 15 min movie before compression
I tend to delete them afterwards just keeping the image series if needed and of course the final compressed movie
Yes, the pointing of your two cameras (your "eyes") has an effect. How much you turn each camera in is called the toe-in. Where the sightlines for the two eyes cross becomes the zero-point of your 3D effect, essentially the plane of the screen. Items behind the toe-in point will appear behind the screen. Items in front will be in front of the screen. The other effect on 3D'ness is the eye-separation - how far apart you make your eyes.
With both cameras simply pointing straight ahead in parallel, you effectively set that zero-point at infinity. Everything will appear to be in front of the screen, which loses out on some of the richness of good 3D when things recede into the distance behind the screen.
Too much toe-in or too much eye separation can lead to eye fatigue, and in some cases key stoning issues. Depending on the toe-in, each eye sees a little bit of a frame that the other eye doesn't. So in your final frames you usually have to trim the opposite sides of each eye to create a frame that overlaps nicely. the pyramid shape you see defining the field of view in a camera is the frustum. The trimming of each eye leads to a slanted pyramid shape called the skewed-frustum. The formulas that tell you how much to trim are fairly straight forward as long as you know the size of your film frame and adjusted focal lengths. Fortunately Daz's camera optics simulate 35mm film (I had to reverse engineer from camera settings and behaviors to double check, but the results seem to confirm that).
When I played around with this, I created my own 3D camera rig. Toyed with idea of creating a product to do this, but it's fairly straight forward so people might not want to pay for it. Maybe I should complete the idea. :-)
So my rig goes something like this... (this might not help you much since you already have 1 eye done, but you might be able to adapt the idea).
Create a camera, this will be my master or "composition" camera. this is what I use to frame my shots, and it's also what I animate. The eye cameras are slaved to this one.
Create a cylinder (invisible in the final renders and no shadow casting!!). Move the origin from Daz's default minimumY to exact centre of the cylinder. With parent-in-place turned off (for all steps in this), make the cylinder a child of the camera and rotated to align with the composition's horizon (probably +90Y), its position is 0,0,0 relative to the camera (same for all child placements in this). The cylinder will control eye-separation.
Create two nulls and child them to the cylinder, each null at one end of the cylinder, These will become the pivots for your cameras.
Create the eye cameras. make them a child of each pivot. Align so they are facing along the Y axis of the pivot null (so +90Y, -90Z) The Y rotation will keep the cameras right-side-up with the composition camera.
Ok, object creation is done. For each of the pivot nulls, set their point-at to the toe-in null.
Move the toe-in along it's Z axis to control the toe-in and where the screen is placed in your 3D view (you could even replace the null with a semi-transparent plane that doesn't render in the final composition?). Change the Y-scale of the cylinder to increase/decrease the eye separation (that's why we moved the origin to the center), toe-in adjusts appropriately.
When I made my rig, I had wrote scripts to slave the eye camera settings to the composition camera (DOF settings, focal length, etc.). Wish I knew more about rigging to do a better job, but it worked.
You could even preview it in Daz by putting left eye on right and right eye on left of two size by side viewports (sized the same). Then use cross-eyed viewing to see the effect.
Render the three cameras and you get your 2D movie (composition camera) and the two eye views. I then run the eye frames through an AviSynth script to trim the edges to create the skewed-frustum and put them in final video frames (I usually made side-by-side frames, you could do top-bottom). I was able to run the video through my LG widescreen 3D TV and it looked pretty darn good.
For normal day to day scenes, commercial 3D studios usually use an eye separation that is slightly less than human average to reduce eye strain -- so around 5 to 6 cm (?? I think human eye range is typically 5.5 to 6.5 cm?) In the old days of slide film and stereo views, the rule of thumb for large scenes was 20:1 ... for every 20 feet of distance to subject, 1 foot of separation. This was "hyper-stereo", more for things like aerial shots of mountains to make them look more 3D in a stereo viewer: e.g. if mountain was 20 miles away, take one shot and then fly "sideways" another 1 mile and take second shot, toeing in so some of mountain was in front of the screen.
See attached image for the setup. The right eye is the red frustum, the left is green. The composition camera is cyan. you can see each frustum extends beyond the composition frame a bit, that's the trimming that needs to be done. The parenting structure is in the inset.
Great, thanks a lot.
Seems I instinctively ended up in a very similar Scenario now.
I fine tuned my Setup like this
one camera object (which I use for the Animation)
two camera objects for the eyes with a little distance which are childs of the first cam
I then created a NULL object and have both eye-cams Point to it.
The result is amazing, didn't think it is that "easy". Now I'm playing a little with a popout effect but I still get a lot of ghosting.
I guess I simply moved my popout too far to the camera, currently recalulating the streams again with much less "popout"...
again: THANKS a lot
possibly. Basically ghosting (or a weird shimmering effect for my eyes :-) ) happens when something is in one eye's view but not the other. If you (or your software) is trimming the edges of the views (the "skewed frustum" mentioned before), then you get this if an object moves too close to the edges and gets trimmed out of one eye view. If you are NOT doing that frame trim, that can lead to some ghosting as well because the final frames don't completely overlap.
General guidelines I've heard from comments from 3D film groups (IMAX, etc.) is they will avoid objects coming too close to the camera (which was a cliché move in the 50's when 3D was popular for a while -- have those spears coming out into the audience so people would instinctively duck). If you do want something to stick out of the screen (into the audience?) then it should be in centre of the frame where the 3D effect is strongest. As it approaches the edges of the frame, it may start to ghost. They try to keep the action in the centre of the frame, and use directorial techniques to keep the audience's eyes on that central area. In a still scene, there's a natural tendency for the eye to want to wander around and out to edges where they will start to strain a bit (because the film's cameras aren't panning with your eyes to move the toe-in point) -- so they use action gimmicks to keep you focused on the center. If they do have wide pastoral/static scenes, then they actually diminish the 3D effect (take off the 3D glasses and you'll see there's almost no separation in the fields) either by putting the toe-in point near infinity, or greatly reducing the eye separation.
another graphic of my rig attached. This is an extreme setup I would never actually use - too much toe-in and too much eye separation, but the extremes show the areas that cause problems.
The composition camera's frame is outlined in cyan. The intersection of the sight lines of the eye cameras shows you where the toe-in point is. That will be what the audience views as the plane of the screen. The composition camera field of view I'm using to show the final frame we would get. If you just combine left and right eyes and align for 3D effect, you wind up with a frame that is bigger than the desired frame.. this is where the trimming comes in I mentioned in that last post. You trim the excess from each eye view that should not be in the final frame (because the other eye can't see it, so you get ghosting). Basically the outside of each eye camera's field of view forms the sides of the frames. The excess of each eye camera on the opposite side is trimmed (so red line shows the area trimmed from right eye's frames, the green line is trimmed from the left eye's frames.
If your toe-in is too strong (cameras turned inwards), you can again enter a situation where there are sections in each eye's view that can't be seen by the other eye, especially as objects approach the camera (think of the eye strain you get when you bring your fingertip towards your nose and you go cross-eyed trying to focus on it) This is especially strong if you've gone with a overly wide eye-separation (hyper-stereo).
So in that case you need to avoid the areas where the eye views are outside your compositions field of view: the shaded yellow areas in the diagram. Any objects that enter that space will also ghost in 3D.
Finally, in cases of strong toe-in -- watch your depth of field effects. Since the cameras are turned inwards, if you apply too shallow a DOF on the eye cameras, the film planes of the eye cameras are at an angle to each other. The edges might start to blur in the eye views. What's sharp on one side for one eye is blurred for the other. When viewed in 3D this causes eye strain and some halo'ing effects. So again you have to coordinate that between the two eye views. Turning on display persistence of the eye cameras in Daz and showing the DOF guides will help you make sure the important areas have similar focus.
If you went with no toe-in (parallel cameras), then you will still get ghosting around the edges and at the areas in-frame that aren't seen seen by both eyes.
Now I just wish I had the rendering power to do large film projects in 3D :-(
Some great information here! Thank you!
So, I presume that the right and left eye cameras are set with the same parameters as the master? Specifically, Frame Width, Focal Length, DOF, Focal Distance, and F/Stop?
I was wondering about the DOF effect that is normally displayed by the master camera in 2D and how is the same effect obtained in the Stereo view? Do you just turn on the DOF in the eye cams?
One last question... if you use StereoMovieMaker, or say Sony Vegas Pro, do you still need to use AviSynth to trim off the overlapping areas prior to importing the clips into these apps?
Thanks again :-)
For my work, the eye cameras are essentially clones of the master composition camera ... so all those settings should be the same. Even with DOF turned on I haven't noticed anything too horrible in the results. I suspect if you did hyperstereo (where the eye cameras are separated by a significantly large amount) and your toe-in point is too close to the camera you might see some DOF funkiness because the film planes are at angles to each other. With "human" eye separation values (I normally work at f/8 to f/11 and 50mm lenses) ...and reasonable "room" like sets/dimensions, most scenes will behave themselves unless an object gets really close to the camera. (remember camera optics 101 --- DOF is affected by focal length, aperture, and distance to subject).
If you take the raw frames into another software package, it depends on if it can do the frustum calculations for you (maybe it takes the separation and depth measurements from you? or maybe it can calculate them on its own by comparing features in the frames?) If you don't do the frame clipping, you will notice the effect as ghosting around the sides of the frame since one eye will see a slice of the scene the that the other one doesn't. If you keep your audience's attention focused at the centre of the frame, you might be able to get away with no clipping and no one will notice.
Fantastic! Thank you! Yep, I just made those changes to my eye cams, and did a render on just a single frame of my scene. It worked great!
StereoMovieMaker and Vegas don't ask for those parameters, so I believe they must perform using information it gathers. I will likely test all different ways to find the results.
Thanks again! :-)
If you ever get interested in other software, iClone has inbuilt stereoscopsy
you can choose red/cyan, side by side (cool but OMG you get headaches watching with crossed eyes!) or two streams the latter for Real3D TV use I presume.
Octane render also has a two colour stereoscopic option.