I recently had some problems rendering print size images in XSI via Mental ray. It all worked out in the end, but it wasn't easy to output the print size renders.
If someone has some time to test out some large image renders in Mantra or has some previous experience i would like to hear about it.
For example try to render some car models (nothing too fancy or heavy poly or texture wise), with basic reflective shaders and some sort of GI, be it AO, FG or whatever Mantra has in the arsenal.
Target resolution: above 8k horizontal.
If this works out without any special image partitioning and especially if it works with polys as subdiv surfaces, I need to save money for Houdini ASAP .
Rendering XXXL image sizes in Mantra
5262 4 1- el_diablo
- Member
- 133 posts
- Joined: July 2005
- Offline
- stevenong
- Member
- 1631 posts
- Joined: July 2005
- Offline
- wolfwood
- Member
- 4262 posts
- Joined: July 2005
- Offline
special image partitioning
Mantra supports crop regions which can be life savers at times. Its not uncommon for *very* memory heavy frames to be rendered in horizontal or vertical strips and then put together afterwards. Have a look at the 4 crop parameters in the camera's View Tab.
The regions should stitch right up.
if(coffees<2,round(float),float)
- el_diablo
- Member
- 133 posts
- Joined: July 2005
- Offline
- Simon
- Member
- 2199 posts
- Joined: July 2005
- Online
Yes we've been happily rendering 6 foot high images at 300 dpi here. (21k)
Always crop large images into your own “buckets” I actually do this using python to modify the ifd, that way you only need to generate 1 ifd and can then split it into as many “buckets” as you like, this also has the benefit of making it easy to distribute the frame onto a renderfarm.
You can still easily do this without python just by using the camera crop settings.
The other thing we have found is that some machines are better than others at doing high res frames, no idea why but we tested all our miscellaneous render boxes and found the ones that never seem to fall over and now we only use them to render big stuff.
Always crop large images into your own “buckets” I actually do this using python to modify the ifd, that way you only need to generate 1 ifd and can then split it into as many “buckets” as you like, this also has the benefit of making it easy to distribute the frame onto a renderfarm.
You can still easily do this without python just by using the camera crop settings.
The other thing we have found is that some machines are better than others at doing high res frames, no idea why but we tested all our miscellaneous render boxes and found the ones that never seem to fall over and now we only use them to render big stuff.
The trick is finding just the right hammer for every screw
-
- Quick Links