Reading through here and looking at what been done it seems like the compositing is taking the approach:
Asset to camera to render to the compositor to render.
Does the data really need to flow in this direction.
What if the final render was built as
Asset to compositor (effects,making and materis) to camera to final render.
This means there is more flexbilty for the user.
It also means the user needs to understand how to hook up 3D renders to look like 3D renders.
Which the community is already very well aware of.
As I understand this, again, I think that those âGMic filtersâ such as exposure, bloom/glare, distortion, etc. should be part of a brand new Camera Nodetree, where you can simulate advanced lens physics, and whatever you expect to be able to control on a real Film Camera.
From there, as it would happen in real world, the formed image is passed to the compositor system, where masking, color grading, and whatever you want can take place.
I know it would be a huge job, at least in terms of workflow and compatibility, but Iâd like to see this happen some time in the future.
i know for a fact that the aftereffects âglowâ filter is one of the most used effects in almost every professional anime production since the mid 2000´s.
of course you can build a glow yourself easily but if it comes at reduced performance or just extra workl its not worth it , professional compositing programs all have their fair share of common effects for easy use
maybe i can find a good example instead , but thats how compositing worked up until now. its much easier to handle masks and layers than it is to handle it in a camera module outside of the compositing workflow.
that would mean , grading , composting, and other post work is still done in compositior but just select things are handled outside of it (in that camera module) , a compositor wants artistic controll over it not realistic handled by camera algorythms. that is too limited to create high aesthetic visuals. movies are all about magik/visual trickery , not about realism. if youve ever been to a movie set, theres so much unrealistic lighting for example. but it looks great once captured to the camera frame(realistic lighting looks like a soap opera from the 90s). same is true in compositing , the bulk of most postproduction is highly artificial and just panders to the aesthetic appeal.
My proposal could live on anyway: everything done in the Camera Nodes would be a pre-composite step. You can do it there or in the actual compositing pass. Both world could live side by side. The plus of having an in-camera composite is (apart the âconceptualâ side of it) that you would be able to have different settings for different cameras. Also dedicated nodes for camera could add features difficult to achieve in other ways (shake node, handheld, etcâŚ)
Yep, Masking and being able to layer everything up is super important. Blender already has a lot of great 3D Tools for 3D space that could be translated into 2D space.
A great example for this kind of conversion would be the Array modifier. A great little bit of kit that has been developed during Covid is a new software for motion graphics called Cavalry. Check it out for free. Cavalry is very similar to Blender and what it can do, but in 2D with a lot more class than Blender.
Modifiers converted to work in 2D composting as nodes is something to think and write home to your mum about. This would allow artists to develop their own lens flares, and bloom, however, they want using nodes. Not having an artist locked into a single node to achieve a remedial effect such as a lens flare, (Rolling my eyes into the back of my head) is very 1990.
Once again to be super clear on this. Masking is incredibly important because without it, your stuck with just Filter @#$%ing(industry term)
These are animation related and we have all the tools to handle them already. Camera nodes could handle optical effets instead, but theyâd be baked into the image and you usually want to avoid that because itâs destructive and best done in comp.
so i played around with the compositor a bit more and its pretty capable already. I made a dirty lens shader which reacts to overall brightness and blends dirty/scratched lens + some bokeh blur, an impact frame shader, my own bloom shader (its not looking as good or is as easy to tweak as the aftereffects glow effect though) and all working in tandem and not even a frame less in playback.
if we get a nice antialiasing module , an âanime pipelineâ will be very achievable.
Sorry to make requests, and I didnât fully read the thread, but here are my thoughts:
Geometry has modifiers
Grease pencil has modifiers and effects
I believe that cameras should have a filter stack, similar to those. And that is where compositing should go. Like, they would have a Filter stack with a bunch of simple filters (Exposure, color correction, blur, etc), or a node-based setup (Compositing nodes) if the user wants to go advanced.
And each camera would have itâs own unique filter stack. It just makes sense to me.
I think that a lot of settings that are in Render and output tabs could be considered to become camera settings instead⌠Like color management, resolution, motion blur (Per camera ISO, Shutter speed, Aperture size), even frame range maybe (So the frame ranges of different cameras can overlap and output different files of different angles for the same frames), so onâŚ
Resolution - most definitely.
Iâve been working on this arch viz project some years ago. Had to render the same model from various angles with many cameras. Some renders had to be one size and aspect ratio, others had to be different⌠I set up each frame to select a different camera, and I could render it all at once overnight with Ctrl+F12 and just go to sleep, but i had to manually input the resolution for each render frame, which really sucked.
An alternative solution would be to animate the resolution, not just camera changes, but resolution animation is disabled.
On âBlender Todayâ, Pablo said that the ability animate resolution was disabled because it would corrupt video files. But I think thatâs a bad solution.
For starters: Itâs highly recommended not to output a video file directly, and render each frame one by one as images, and connect them later. So, why have video output as an option at all? Instead, there could be a simple built-in tool to connect image sequencesâŚ
But if thatâs not an option, the videos wouldnât get corrupted if the resolution differences between frames are fixed with black bars.
And yet, a simple warning text for animated resolutions could be enough⌠The inability to animate resolution is a far bigger issue than the edge case of possibly getting a corrupted videoâŚ
if( output type is one of [all video formats] AND scene resolution has keyframes)
then {
show_user_warning(
"Output Video Types are incompatible with animated scene resolution values.
Videos need all frames to be exactly the same size"
) }
⌠however⌠Iâve used Lossless Cut to join together a dozen different video files of wildly different sizes and formats.
VLC on my laptops can play them just fine. Doesnât work in VLC on phones/tablets/amazon firesticks.