Healthy debate question…
Why would we want SideFX to invest money into revamping COPS?
For compositioning - there’s Nuke and others already…
For materials - we have materialsX and VOPs
For complex materials - there’s substance designer…

I keep hearing that revamped COPS would be a Substance designer killer but how?
What could Houdini offer that other don’t?
Can we not already do what Designer does - in VOPS?

Craving your perspectives!

• context fluidity. the interoperability of cops with the rest of houdini (when it works) can't be matched by using external software. writing a cops raytracer, having heightfields directly sample into textures, directly convert image sequence cops into volumes, you can't do that with a combo of houdini and nuke and substance. If these cops workflows could be made reliable, hoo boy.
  
• realtime needs. its a specific subset of the above, but games and gpus will require baking whatever sort of information down to data textures for a while. again this is clunky if your 3d app can't maniplate textures effectively, it'd be a pain to make this work cleanly by writing out some intermediate format from houdini, to pull into nuke/substance. the labs games tools only work because of the abilities of cops, but it could be so much better with even tighter integration.
  
• cost. nuke is expensive, substance designer can get expensive, houdini core is cheap by comparison if you can make it do all the things eventually.
  
• workflow for lighting. from what i understand the orginal design of katana was to have a basic compositor within it for bash comps, cg lighting is largely A over B with grades. its surprising but understandable that the foundry removed this when they pushed katana to a commercial product (they gotta push nuke seats after all), but proper useful bash comp tools within solaris would be huge
  
• the unknowns. much like kinefx was targeted at mocap editing at first but folk have found all kinds of crazy fun to be had building flower and tree rigs, tentacle rigs, rbd editing tricks, i wager an update to a compositing framework within hoduini would unlock all sorts of ideas we haven't thought of.
  

Even with vops, yes on paper it does a lot of what substance does, but in practice its not really. All the pixel operations you'd expect are lacking (blurs, convolves etc), its not straightforward to do image exports, nor straightforward to do layering operations or branch outs like what substance can do for diffuse/roughness/occlusion/normals.

Other existing tools shouldn't be a showstopper at all. Houdini has improved its modeling toolset over the last few years despite plenty of efficient modeling tools out there. A lot of work is put into a GPU renderer when many external GPU renderers already exist. If you don't develop because another option already exists you might as well stop entirely.

From what I understand, Nuke doesn't have the equivalent of a VOPCOP, so there's one thing that COPs brings to the table even in its current state. A while back I had a need for an animated pixelation effect that has different sized pixels top to bottom and gradually refines itself: using a VOPCOP I put it together very quickly.

Clearly, revamped COPs would unlock much more power than just compositing. Of course we want it to be able to layer a bunch of passes together, but I see its main uses in things like texture synthesis, custom effects and direct interoperability with other parts of the application.

Was it known that SideFX was working their "next generation, procedural image processing system"?

There is a job posting for a compositing software developer which states similar things matt and digipiction have touch upon:

We are developing a next generation, procedural image processing system to provide the backbone for traditional compositing, texture synthesis, realtime 3D preview, and motion tracking. You will be working to integrate these capabilities within Houdini, our flagships software, which has proceduralism, openness, and flexibility as its core philosophies. The new compositing system will embody these ideals, scaling to massive task graphs, supporting both out-of-the-box operations and user-generated algorithms. It will work seamlessly with other parts of Houdini, including the 3D viewer, the Solaris/USD environment, our Karma renderers, physical simulations, and geometry processing.

Hi. Can you image the new system with PDG added in. There is nothing like PDG in any other app I have used. Best

I have been playing around with COPs a good bit recently, mostly for texture generation. I have beeninterested in trying to recreate some substance designer type textures in Houdini. COPs has many quirks,however, it really is quite flexible for an old system. Even with all the quirks I found it relatively easy to rebuild alot of substance designer functionality fairly easily.

To mention some quick pros for your substance question.
In substance we can spend alot of time trying to define our base geometric shapes.
We can build lots of complex shapes in SOPs and with Heightfields that can create interesting patterns quickly. Importing geo attribute into COPs also has lots of potential. The main component it lacks by default is easy ways to tile textures, which I think is solvable.

Some cons that have stuck out that are worth noting,
The channel copy UI is poor. Setting the source and then the target feels somehow inverted to me, though you can of course VOPCOP your way around this.
The composite node can be confusing. I often dont get the result Im expecting from the operations. While this is probably because Im not a great compositor, it feels like results that I would expect with a merge in Nuke or even thinking back to my Shake days do not work for me in COPs. For example when fading down a layer (mixing in Nuke) often results in black images rather than a fade and Im unsure whether to use the foreground/background mix or the masking effect mix.

I mention these two things in particular as the composite node really is the backbone of COPs and I think that channel copying (shuffling) is another important concept where new COPS users will run into a brick wall pretty quickly. Along with instability I can imagine it been a frustrating introduction to compositing in Houdini for most users, which forces sidefx to make a decision to kill it or embrace it. I have some simple intro tips over on Youtube to try and get new users over some of the humps. You can find them here https://www.youtube.com/channel/UCu8pjcmyzDO9iLmCuwsv6uw [www.youtube.com]

but it feels like alot of this needs to be addressed at a UI/UX level.


Outside of my recent interest in substance style texturing, it could be quite appealing for game artists both for texture generation and also fx texture creation. Some basic camera tracking functionality could be useful for smaller teams. For lighters it allows for automated slap comps as well as possibly building texture generation into their look development workflows.

Beyond this I think that we will see more 2D information being useful for 3d, such as 2d motion capture for characters both body and face, as well as interesting developments in AI image generation for example Im having lots of fun lately running estimated depth maps through COPs. Needs some serious investment but potentially worth it imho.

LukeP
Healthy debate question…
Why would we want SideFX to invest money into revamping COPS?

I believe the primary reason for wanting a revamp is that COPs is highly unstable and slow. These reasons alone are significant enough to justify a revamp, to me.
Additionally, when considering the other valid points mentioned earlier, it becomes clear that something needs to be done about this situation.