merging scenes in same location

[deepblue812]

Been trying out Lavina (0.5) last couple days and am amazed that it takes pretty much whatever scene I can throw at it

As a test, I'm trying to save out an animated drive-through of one of my larger files that consists of an airport and garage with several cameras that drive around the different areas (garage, arrivals, departures, etc)

While I can save out the scene from 3dsMax as a .vrscene with animation turned on, it's quite slow for three, 3000 frame cameras, so I tried the idea of just saving out the main model as a .vrscene, then creating an animated .vrscene of just each camera with my 'driver view' car, which obviously saves out very quickly, then merging them in Lavina to capture whichever camera I needed. This does seem to work, but the problem is the scenes are coming in at very different locations.

I thought maybe Lavina was trying to 'center' the mesh on opening/merge so I put a small object at 0,0,0 in each scene to see if it would align the scenes at a common point. That didn't work, so I tried putting several objects around the extents of the scenes to create something of a common bounding box but that didn't work either.

Seems like it should be easy, so am thinking I'm just missing something.

Thanks

SJ

[Phillip Miller]

Thanks for the feedback SJ.

We will definitely be including an option with Merge that will insert your chosen scene according to the origin without any placement.
In the meantime when using Merge, try pointing your cursor "into the sky" where there is no geometry - then the merge feature has nothing to hit-test with and placement should be according to the origin.

Kind regards,

- Phil

[npg]

You said adding an object at 0,0,0 didn't work, but what does this mean? The central object didn't show up where your cursor was (where the red arrow indicator was)? The red arrow indicator is aligned with the normal at the intersection point and shows where the merged scene's "UP" will point to. You can make it match the global "UP" direction instead by holding Shift while dragging.

The bottom of the bounding box of the scene being dropped is aligned at the drop position (sideways translation is done as well, such that the center of the box is over the drop point). Maybe this is what caused the surprising behavior? The idea was this would be used for small assets such as a tree or a flower pot or a painting, aligned to another object (wall, ground). The feature is experimental and we obviously need to extend it. We probably need to add a dialog where you can specify how exactly to align the scene, etc. What options would you like to see?

By the way, the reason we don't preview the scene being dropped (or at least a bounding box) is that it can be arbitrarily large and complex and the vrscene format does not include an easily accessible bounding box. We would have to process the whole scene even just to get its extents

[deepblue812]

Thanks, that makes sense. This is definitely a lot more than dropping a teapot next to a tree
In the spirit of pushing it till it breaks I was essentially merging an entire airport model into a scene with 4000 frames of animated freeway and vehicles. Good news is that after some tinkering it pretty much completely worked.
I saw the little 'drop helper' arrow, but as Phil mentioned dropping the scene off to the side so it could assume it's own coordinate system seemed to work after some trial and error.

Trick was that I had to leave the camera in both exported vrscene files even though I only needed the camera animated for one of them to create the drivethrough path.

So the main file becomes the vrscene saved out as animated with just the camera and moving vehicles, the file that gets merged into that is the bulk of the building and roadway geometry exported from and with that same camera. Originally I had been trying export the geometry without the camera in the scene and it wasn't merging in the correct location (shifted about 500' in the scene). After I exported the static geometry with the camera included it merged in lined up with the animated camera.

It probably would have worked as exporting out the vrscene entirely just from the animated camera, but would have taken a long, long time to export so was looking for a way to break up the animated parts from the static geometry.

The only thing I noticed that didn't work is some ForestPro grass/tree objects. They kept the incorrect 'shift' when merging the static vrscene into the main Lavina scene with the animation. If I opened (not merged) the static scene just by itself the ForestPro objects were in the correct spot. I might try exporting the ForestPro objects with the animated camera as a test which I think would work since my guess is it would 'lock' those objects to the camera view.

Not sure what all that means at the end of the day for file merging workflow. Just a basic dialog with some coordinate awareness would probably be fine, being able to name some kind of common point between vrscene files I think would work. Without the same camera in both files things wouldn't stay lined up even though everything was being exported from the same file. Will need to test more on some other scenes to see if there's different behavior. Could just be 3dsMax messing things up and Lavina just trying to do what it thinks Max wants to do. The files I'm using are certainly not built around real-time workflow in mind, just happened to be the project I was working on at the moment and seemed like a good test to throw at Lavina. So far very impressed!

SJ