Multiple cameras, one cook?

[meldrew]

Hi all,

Ok, so I have a scene/set of scenes where I need to render the same sequence from numerous cameras.

Is there a way to render, say, 4 cameras at the same position, but facing different directions, but only calculating the lighting/reflections/shading etc. once?

As far as I understand .ifd just include the instructions to render, not an actual cache of the scene itself. 

Time is a little tight, so any optimisation would be a benefit.

(Cams are also various resolutions, not uniform, in case that has an impact.)

Thanks!

[lukeiamyourfather]

45 minutes ago, meldrew said:

Is there a way to render, say, 4 cameras at the same position, but facing different directions, but only calculating the lighting/reflections/shading etc. once?

Mantra renders only what it needs to on a per pixel level so there wouldn't be any savings in terms of calculations like that.

46 minutes ago, meldrew said:

As far as I understand .ifd just include the instructions to render, not an actual cache of the scene itself. 

The IFD file could potentially contain the entire scene but it can also use existing geometry on disk that's already cached out. This is where you'll find the most potential for optimization. If you can cache as much as possible before the IFD stage then there's less duplicated data (less disk space required, less network throughput required, etc.). If your IFD folder is more than a hundred megabytes in total then there's room for optimization here. Use File nodes that load packed geometry and Mantra will load them directly rather than stuffing them into the IFD.

It sounds like you might be trying to render out a cube map or something like that. Mantra has several cameras for VR and surround rendering. If that's what you're doing you might want to check those out.

[meldrew]

5 minutes ago, lukeiamyourfather said:

Mantra renders only what it needs to on a per pixel level so there wouldn't be any savings in terms of calculations like that.

Hi Luke, thanks for the quick response  - So if I understand you correctly, Mantra isn't thinking about rendering anything outside the frustum in any case? If so that's great (and what I assumed/hoped anyway!)

 

7 minutes ago, lukeiamyourfather said:

If you can cache as much as possible before the IFD stage then there's less duplicated data

I assume you mean saving geo out as .bgeo, caching sims etc.? Then pointing IFD directly to these caches? I'll need to look more into this & packed geo, as I'm rendering on a cloud service... But that is another thread I suppose. ha.

I always endeavour to cache things out as efficiently as I can, so I'll revisit this and see if there's any improvements I can make.

 

8 minutes ago, lukeiamyourfather said:

It sounds like you might be trying to render out a cube map or something like that. Mantra has several cameras for VR and surround rendering. If that's what you're doing you might want to check those out.

It is a cube map of sorts yes - However it's not for VR, so any spherical mapping/fish-eye lens type solutions will lead to distortion that is not wanted in this case.

I'm rendering content for 3 walls & floor of a box, all angles  90deg, with the pov at eye height - roughly centred within said box. Would there be a specific camera you would recommend? 

 

Thanks in advance for any additional pointers, appreciate your time. I know these are fairly rudimentary questions