This frame is from an animation I recently generated using Blender. I used my csv_animate Python script to import motion data into Blender to drive the animation. A second script generated camera position and orientation to track the vehicle. I’ve posted regarding csv_animate in other forums, and you can find it here
The model geometry was imported from CAD models, then materials applied and rigging done in Blender. (Thanks to Greybeard for a tip on Object Hooks.)
I compiled Blender with a change to allow a negative Zoffset to be used to trigger rendering an object with maximum z-buffer depth, forcing that object behind all others. For details see:
This was used on the ground plane to avoid issues with z-buffer quantization that were causing objects on the ground to intermittently disappear behind the ground plane. That’s what’s great about open source software–if you need a change, you can make it. In hindsight I might have been able to use the Sequencer to do the same thing. (This is the first time I’ve used Blender.)
Some tricks I used to speed up rendering (for my fellow noobs):
- For surfaces that have image map textures, especially the ground, make sure “mipmap” is on. This will cause a delay for the first frame rendered while Blender creates a mipmap, but then per-frame times are much shorter afterward because anti-aliasing has been done in advance.
- For groundplane, sky and other non-vital objects, set materials to “Shadeless” and disable “Traceable” and “Shadow”. This disables raytracing and shadow casting for these objects for a significant speed up. Shadows of objects on the ground were faked with black polygons.
Faces: 85K (does that qualify as hi-poly?
1 hemi blue tinged pointed down to simulate sky light
1 hemi yellow tinged pointed up to simulate light reflected from the ground
Memory used during render: 500 MB (largely due to the ground plane texture image and mipmap)