Mash

I just posted the news article in our work chat, not one of us had ever heard of them before

Zero sales, everyone was busy playing Mario kart.

This is what the fracture object (not voronoi) is for. it allows you to custom make each mograph element

Youre creating a volumetric system which is only a single voxel thick. This means when your diagonal curve ceiling leaves one voxel and enters another, youre going to have a gap as some polys are too far away from both voxels to be considered as belonging to them. Its the 3D version of having an aliasing diagonal line Honestly, you're asking a lot from the system trying to burn down an entire intricate interior. You either need to bump up the thickness/emission of the burning areas so the lines get filled in; use a higher res voxel mesh or just find another way of doing this. My instinct here would be to start off with a base of simple animated burning elements with alphas, then comp in extra smoke or more detailed areas as needed.

It depends what you're looking for. Standard and physical can give you absolutely lightning fast renders, on a modern machine you can be doing 2-3 seconds per frame. On condition that you don't care about GI, AO, motion blur, DOF, frosted glass or mirrors etc. For simple illustrative work they are rock solid and fast. They're also pretty good with the sketch and toon system, though it is showing its age and they did ruin it a while back with the AA overhaul. Similarly the hair can look really nice, so long as you're not fussed about accurate GI etc. If you want real world modern lighting though, forget about it, theyre slow and have way too many drawbacks once you start flipping on all the switches. Personally I use octane for my daily work. Its absurdly fast, gives great lighting control and I can have all the bells and whistles enabled, I know the dof will be beautiful every time, the motion blur will be fine. The key thing is that its very quick to develop a look with its realtime feedback. That said, its terrible for NPR and it crashes like a dick multiple times a day. I feel like the 3ds max users we used to mock back in the day. Regarding architectural work, it isn't terrible, many of us used it fine for years on end, but I'd never pick t for that these days. The GI is slow, grainy and flicker ridden. Half our work was trying to fake GI so we could avoid using it.

one problem is that your base mesh isnt actually flat. If you crank the phong shading angle down to 0 and stick a light to the side, you can see the middle is flat, but the curves surfaces have depth to them: If I just run a plane through the base mesh, you can see these parts all stick out

Just because the control mesh is straight, doesnt mean the resulting SDS mesh will be smooth. The different heights are going to cause different rates of curvature top to bottom. In this case I would go with either a volume builder based approach, or just leave the edges sharp and use material rounding to smoothen the edges volume.zip

... and on that, is there a way to get this properly into photoshop? PS has no 'interpret' options for me to discard the alpha to do the same trick I do in AE, outside of rendering the image again with no alpha channel. Strictly speaking I CAN get the content into PS and it does look good if I go via 32bit exr, but once in, I cant squash it down to 8 or 16 bit as doing this re-obliterates the glass again.

Skip to 3:20

Ok, so my current method of working in AE where I take a copy of the render interpreted with no alpha channel (so I get all the rgb data) and set it to add mode; then use a second copy of the render with the alpha enabled and throw it on top as a solid normal mode is pretty much correct then? ie. its as good as adobe AE is going to get given its shortcomings.

Then pardon my daft question, but what would be the correct way to composite this in post? Lets say you have a red glowing cube you want to bring in on top of your composition. You bring in the cube renders (premultiplied so the glow is maintained in the rgb data), they sit on top, but the alpha layer outside of the cube in every 3d render engine ive used will be solid black because there is no geometry there, thus the glow of the red cube is completely cut off. Previously (and currently frankly) I have been bringing in a second copy of the footage (which does have a glow visible in it, but the alpha obliterates it), placing it under the red cube layer, disabling the alpha entirely, then setting it to add or screen so the glow can show back up again and composite over my background. But this always needs 2 copies of the render in my comp, if I just use the add/screen layer then the red cube would show the background through it.

I think you may have this backwards. In plain language: Straight alpha: Stores the full RGB colour for each pixel, but ignores how transparent it may or may not be. ie the transparency of a pixel has no impact on the colour stored. This means for example if you render a white cloud with a soft whispy edge in a blue sky, the rendered cloud will only contain white cloud colours, the blue of the sky will not be present in the rendered cloud, even where the alpha transparency eats into the cloud. Premultiplied: This simply means the image being rendered has already had the background baked into the rendered colour data. In the cloud example it means that it will start to turn blue as the edge of the cloud becomes more transparent. In practical terms, straight alphas can be great because there's no bleeding of the background into the visual RGB data, you can take your white cloud and throw it onto any background you like, there wont be any blue from the sky creeping in. On the other hand If you place your premultiplied cloud onto an orange sunset background, youll get a blue halo around the cloud, which sucks. However.... It isn't all roses. Sometimes you need the background colour to be baked into the transparent edge because some things are just flat out impossible to render due to the number of layers present or the motion on screen. Here's one which screws me over regularly; what happens if I have a 100% transparent plastic fan blade, but the fan blade is frosted. And in the middle of the fan is a bright red light. Visually the fan blade has the effect of looking like a swooshing darth vader light sabre. Its bright red and burning white from the brightness, but whats there? a 100% transparent object.... The alpha channel with a straight alpha will obliterate my rendering, its 100% transparent plastic. You can see it, but the alpha channel decides to ruin your day and now the rendering is useless. The only option here is a premultiplied alpha where the background is baked into the motion blur and SSS of the plastic fan blade. Sure, I need to make sure my 3d background somewhat matches my intended compositing background, but its the only way to get any sort of useful render. Same goes for motion blur, DOF blur, multiple transparent layers in front of each other (steam behind glass) The honest answer is, use whichever one is least likely to screw you over. If you have lots of annoying transparent/blurry things to deal with, go premultiplied but plan your background ahead of time. If you want clean alphas on a simpler object render, go straight alpha. I haven't read your linked blog all the way through, but I will say... there are an abundance of wrong people loudly proclaiming themselves to be fonts of all knowledge. Theres one in the octane community who insists on inserting himself into literally every thread on the entire octane forum to tell you youre an idiot for using a png file, he has 100's of blog pages which are a strange mix between 3d rendering and flat earth magical woo woo maths to show everyone just how right he is. That said, your rainbow example does match up with what the blog says. The only difference is the blog seems to think the straight alpha is evil and you should only use the premultiplied, whilst I would say both have their uses, with straight being preferable when possible.

If I were in this situation, I would disconnect my machine from the internet (kill wifi, pull ethernet), because then theres a good chance it will simply timeout after 30 seconds.

Perfectly reasonable price for what it does tbh. If you need to make a minecraft world environment for a job then its an absolute bargain; and with the release of the film its probably a pretty good time to release it.

Basically yes. It forces c4d to run the crash routines which includes attempting to save all open files into the bug reports folder. Theres a slightly different key combo on mac, ctrl alt cmd backspace i believe.