[Steve Bentley]

So basically you want to see the wood grain all the way through as you would with a boolean on a solid.

[Steve Bentley]

I’m not sure I understand what you are after. Do you mean when you do your particle effect in Maya or C4d, the shell of the model gets eaten away as the particles(polygons) move away, leaving you with a half eaten kinder egg who’s shell keeps being eaten more over time? In other words you can see inside it?
Are you looking for more of a alpha channel reveal? This can certainly be done right in AE or you could make elements for this in either maya or C4d.
Do you have an example of the kind of look you are going for?

[Steve Bentley]

I can see a few issues with this process I thought I’d mention before you head down the rabbit hole:

The models coming from different packages will have different mapping associated with them, so unless you are going to do a plain jane material and let it wrap however it likes, you may end up with strange looking textures.

Are all the models the same size? (including the same scale and same orientation?) Some packages work in different units to C4d and some even have Z as up or Y inverted, so your models may come in too big or backwards or upside down.
And then there’s the camera to frame the model’s different sizes – you will have to include math to get world width and height and work out how far back the camera should be for each model, based on focal length and aspect ratio.

This doesn’t solve any of the above but is a hack for loading. There are mesh loaders out there (like the one that comes with Real Flo – C4D may have one by now). If you have OBJs and name them all sequentially, C4d will load them, one for every frame. Then just render a movie, however many frames long you have models for and you’ve got all your shots.

[Steve Bentley]

I wouldn’t say its not important. It really depends on what you are doing. The bigger you can go the better experience you are going to have. If you are doing logo animations, a bigger card will make a bigger difference. If you are doing 100+ layer deep effects shots, while it will help, you won’t see the kind of difference you would with graphic based elements. If you preRender elements and then do your big composite with those prerendered elements, making those prerenders will run very fast, so again a bigger difference with this work flow. If you keep all your layers live with effects intact and have long timelines, a big card will have less of an impact.

The old adage still applies- get the biggest and fastest you can afford – once you drive that graphics card off the lot, it becomes obsolete.
I don’t have a history with AMD/ATI cards in boxes (although, ironically we used to animate Ruby for them and do box art – with Nvidia’s), but if you polled everyone here, I’ll bet most would be Nvidia where AE is concerned. It may also be platform dependent. You may have a better time with one or the other on a specific platform. Do Mac’s generally come with ATI or Nvidia? (We switched over a while ago so I’m not current on Mac architecture). If Apple prefers one brand over the other (they’d make their own by now if Jobs was still with us), they may have built their machines with a particular card in mind.

[Steve Bentley]

I just stumbled on this and thought I would update and correct it. The syntax for the above is incorrect in both posts. It should be
thisComp.layer(index+1).position;

[Steve Bentley]

We’re routinely doing 20k pixel wide comps with only 4GB of vram on some machines. The thing is that you can only pack so much ram on a video card (at least in a sub $10k price range) and with larger comps there just won’t be enough, period, so to fill the gap the graphic card starts chomping up system ram as its designed to.

While there are certain things that graphics cards do faster than what the system ram can do, there are so many things that glitch with AE and GPUs or have different looks with GPUs that in the end we find system ram more important for large files. And with any kind of serious compositing you are basically running on system ram anyway because the video ram got filled up on that first frame it rendered.
I’ve done 20k files with 16gb of system ram but its slow (lots of disk writes). So it is possible to do big comps with low levels in both system and vram, although you just can’t go back once you jack the system to 128gb and jam a Titan XP in there. But even then the Titan only has 12gb of Vram (I know I know, its the new RX5, still…)

[Steve Bentley]

Just so you can see what you are getting into, here’s how it normally done:
Assets:

3D model of rubics cube.
3d model or scan of the persons hand (low poly hand may be enough – whether you need this element or not depends on the shot)
High quality hand held footage of actor with empty hand (usually with tracking markers). (re high quality: the larger and closer to uncompressed the footage is the better the track)

Method
3D track the hand and the world.
Rotoscope the fingers that cover parts of the cube and make a matte.
Remove the tracking dots from the live action hand.
Bring the camera solve from the 3D tracker into the 3D package and set up your “world” to include the 3D hand and rubics cube and correct lighting. Generate a reflection map that corresponds to the world and the hand.
Match move the 3d hand model to the actors movement. This may involve more than just matching the motion, often you will have to animate the hand (finger flexing etc) to match.
Attach the movement of the hand to the rubics cube and then animate the rubics cube if it does more than just sit in the hand. Run a dynamics simulation (or animate by hand) to have the rubics cube fall apart, with the sub cubes of the larger cube interacting with each other – this requires a physics engine and rigid body simulations and some gravity.
Turn off the 3D hand and render 3D rubics cube element with alpha channel, depth of field channel, motion blur channel and a shadow pass channel.
Composite the 3D elements onto the 2d shot, using the rotoscoped element to mask out obscured parts of the cube in the hand and the shadow pass element. Add a little frame by frame hand paint to fix anything that needs it (its gonna happen)
Color correct elements so they match.

Adding a 3D element into a 2D shot is hard but gets exponentially harder when it has to be: in contact with a human, a moving human, a moving human and a moving camera.
A few years ago, we(and a few other companies) were asked to replace a baby doll in a movie with either a CG baby or a real one, where the main actor holds and cradles the doll. There was no amount of money they could have offered where we would have said yes. The baby doll remains in the movie. If elements had been shot to accommodate such an effect it would have still been hard, but as it was…..

What you need to do is very doable, but its not for the faint of heart, assuming it has to look real.

[Steve Bentley]

It looks like that red line is just the outline for one of the layers. AE shows you this so footage that extends beyond the frame can positioned even if it doesn’t show up in the viewport. Only the stuff withing the stage/visible frame will render.

[Steve Bentley]

Stupid question but have you chosen connect all objects? Are all the nulls in the tracer link box (you can’t just put the parent null in there).
Even with them inside the tracker you should be able to put the “child” nulls of the tracker object inside the link box of the tracer.
Worst case you could make a list of nulls out of the ones in the tracker and link those to a list of stunt nulls in expresso and use the stunt nulls in the tracer link box.

[Steve Bentley]

So to ask a dumb question – is there a reason to render out at 32bit (for your project?) Is it possible the container you rendering out to doesn’t support 32bit? I know that they sometimes call quicktime with alpha “32bit” but that’s total number of bits across all channels (including an alpha) not 32 bits per channel.
Do I have that right? you were trying to render TO a 32 bit file?