Hello Houdini folks

Do we really need Nuke or After Effects for Compositing work ? or you think that Houdini compositor tool can do all things without need to buy/Learn new compositor software ?

COPs is nice…but there is no way that it could replace Nuke in production…

thanks arctor , You've been very helpful Person

thanks again

LiveCycle
thanks arctor , You've been very helpful Person

thanks again

COPs is also nice for doing precomps and procedural post processing effects since you have access to your scene within your hip file. But if your task is to comp day after day, Nuke all the way.

it's very beneficial to learn COPs since they have some great features you may not find in other compositing packages especially VOP COPs, where you can make your own filters which communicate with other contexts, very powerful stuff.

But just look at the COP nodes and you can guess by yourself to what extent it can be useful for your needs, if you are compositor, you will pretty quickly discover that majority of useful tools are missing like, tracker, paint, optical flow retiming, proper stereo tools, proper keyers and many others
so as others said, it's not as mature as major compositing packages, still very useful though

COPS is an indispensable part of Houdini, especially for image processing that is meant to feed your other networks, and in fact I wish Side Effects would turn their attention to it a bit more and spruce a few things up so I could use it more often to finish off renders.

That said, there is no way it can replace Nuke, and you will never find that happening in a studio. Nuke is just too good at what it does.

Any further thoughts on this topic since 2012? Referencing COPS seems a lot more stable since the introduction of PDG.

Michela Ledwidge
Any further thoughts on this topic since 2012? Referencing COPS seems a lot more stable since the introduction of PDG.

COPs is more stable than it was but it is very slow, especially reading images. There are a lot of good ideas in COPs but it's been neglected for a long time now. Whenever I try to do something half serious with it I still discover bugs. It would be nice to have it turn into some kind of lookdev tool, like substance. Ah, one can dream.

Same with mplay btw, still good but a few more years of stagnation and we might consider it outdated.

Michela Ledwidge
Any further thoughts on this topic since 2012?

Being a one-man-army, for me the idea of using an external compositor is a key part of the creative process. Blender has a more useful compositing page that is quite fast, and yet I think even that is no replacement for a proper compositing pass.

First of all, I leverage third party plugins all the time to generate film grain, lens flares, glow effects, blurring, and even the surrounding environment. Trying to replicate this in Houdini would be like trying to drive a nail into a brick with the tip of a screwdriver -- wrong tool for the job.

Secondly, I find that switching to a completely different app resets my visual brain and allows me to look at the shot in a different context which in turn helps me consider ways to improve it.

My typical workflow pipeline is Houdini render AOVs with 3Delight --> Aftereffects or Fusion (depending on the needs of the shot) --> Davinci Resolve for color grade and final output.

Probably still futile using it for tracking, roto and keying.
However with vex you get total control of the image without the risks that come with with glsl/gpu.
Having said that I'm really liking what I see with Tooll3 so-far...

Midphase
First of all, I leverage third party plugins all the time to generate film grain, lens flares, glow effects, blurring, and even the surrounding environment. Trying to replicate this in Houdini would be like trying to drive a nail into a brick with the tip of a screwdriver -- wrong tool for the job.

If COPs were faster (please please make it faster!) I would have to strongly disagree with you. The crazy thing about COPs is that it gives you all the same power that SOPs does. You can literally use the same VOP nodes that you use in SOPs/MATs to produce extremely complex effects.

I've used VOP COPs to create custom blurs, custom grain, a ramped pixelization effect and many other things. You can turn all of this into HDAs like you would in other contexts. Time and time again I'm disappointed that it's so powerful but so slow.

Among other things we currently use COPs to output per-vertex data into EXR files: writing vertex 1 to the pixel at the bottom left. Vertex 2 to the pixel next to it and so on. Highly custom stuff, which I assume no After Effects plugin exists for.

It strikes me that there's both tech opportunities and market opportunities to justify work on cops sooner rather than later.

-Games! Games and realtime are using textures as data storage more than ever, cops via VAT and other tools is a great way to author these things, but it's sitting on tech that has barely changed in 10 years. Imagine all the possibilities if cops had kept pace with the rest of houdini.
-Substance! Substance designer shows how powerful a node based texture system can be, but the Adobe stuff of late is generating a lot of bad vibes, and the little bits I've played with imply a pretty bad workflow. I don't expect cops to ever compete with Designer at the super highend, but most of the common operations designer is used for should be possible, combined with the ability to read and write data from sops in a way designer will never be able to fully match.
-Film/VFX comping! Nuke is still the workhorse for film and vfx, but it's expensive, stagnating, and often overkill for what most people use it for (bash comps and pass layering). Cops is almost good enough for these kind of tasks, but sticks on a few key things that means it can't be relied on (Tomas touched on a lot of these earlier like roto, keying, speed, reliability)
-Bash comps and Solaris! With Solaris pushing into the lighting scene, the need for a good bash comp tool is even more urgent. If studios worked out that for the cost of a Core licence they could replace both a Katana and Nuke seat, that could be huge.
-OpenCL! The guts of the heightfield nodes intrigue me as a possible tech base for a new cops system. That heightfields can throw around pretty large data sets so quickly, and that there's a direct interchange to cops and back, and seeing the kind of speeds Resolve gets with its gpu accellerated compositing. Very exciting.

There are currently a lot of missed opportunities with COPS, whilst I wouldn't really advocate it to be your be-all compositing solution it 100% SHOULD be your solution for image manipulation in the context of houdini.

The tools are there, the proceduralism is there, TOPS are there, but the speed and stability is not, which is such a waste of potential. I'm now using cops basically every day for a variety of different things, from simple render cleanups, denoising, the labs baking tools, and all of this hooked in with TOPS (go on, try to make me call it pdg...) makes for an extremely powerful and flexible workflow but the COPS side of it is just so slow and unruly.

I do wish sidefx would spend some of their resources on something that would be a (literal) game changer with COPS 3.0

My dos cents.

mestela
It strikes me that there's both tech opportunities and market opportunities to justify work on cops sooner rather than later.

-Games! Games and realtime are using textures as data storage more than ever, cops via VAT and other tools is a great way to author these things, but it's sitting on tech that has barely changed in 10 years. Imagine all the possibilities if cops had kept pace with the rest of houdini.
-Substance! Substance designer shows how powerful a node based texture system can be, but the Adobe stuff of late is generating a lot of bad vibes, and the little bits I've played with imply a pretty bad workflow. I don't expect cops to ever compete with Designer at the super highend, but most of the common operations designer is used for should be possible, combined with the ability to read and write data from sops in a way designer will never be able to fully match.
-Film/VFX comping! Nuke is still the workhorse for film and vfx, but it's expensive, stagnating, and often overkill for what most people use it for (bash comps and pass layering). Cops is almost good enough for these kind of tasks, but sticks on a few key things that means it can't be relied on (Tomas touched on a lot of these earlier like roto, keying, speed, reliability)
-Bash comps and Solaris! With Solaris pushing into the lighting scene, the need for a good bash comp tool is even more urgent. If studios worked out that for the cost of a Core licence they could replace both a Katana and Nuke seat, that could be huge.
-OpenCL! The guts of the heightfield nodes intrigue me as a possible tech base for a new cops system. That heightfields can throw around pretty large data sets so quickly, and that there's a direct interchange to cops and back, and seeing the kind of speeds Resolve gets with its gpu accellerated compositing. Very exciting.

Could not agree more, well said!

You are all forgetting what COPs is missing the most, and that's accessibility. The available information for beginners to even get started with this is pretty much non-existent. I have had multiple forays into COPs but have abandoned all of them because of how inscrutable it is. The same goes for CHOPs, btw.

With the exception of a couple of very specific things I do with with COPs and CHOPs, it's unreasonably difficult to get into either of them.

If SideFX is going to give either of these any love, I wish they'd also ramp up the learning material for them, like they have been with other toolsets.

eikonoklastes
You are all forgetting what COPs is missing the most, and that's accessibility. The available information for beginners to even get started with this is pretty much non-existent. I have had multiple forays into COPs but have abandoned all of them because of how inscrutable it is. The same goes for CHOPs, btw.

It's true that there's not a lot of information out there. If there were more learning materials, people would certainly be more inclined to give it a try. And Side Effects might get incentivized to modernize COPS.

Still, I've found it not too difficult to get into COPs because it's quite similar to SOPs and VOPs. CHOPs on the other hand is dark magic to me, I don't touch it because I have no clue how it works.

COPS is the one area of Houdini I really wish would see a significant improvement sooner rather than later.

I pretty much use it every other day, and since the introduction of TOPS it has become even more indispensable for our workflows. Sure, I use to export things to AE, Nuke or Substance for all kinds of manipulations, but since I switch to working on games the need for custom image work for data storage as mentioned above by @GoetzingerC and @mestela has been one of the main things I use it for. And not having to jump to an external editor is very productive.

It's just a shame it's so damn slow. It's the perfect complement for TOPS and reduces the need for custom studio scripts that were poorly written years ago for simple things like: render, comp, export.

One can dream and I hope there's some love for COPS sometime soon <3

I personally experienced a ton of crashes while attempting to use COPs, segmentation faults left right and center, I needed to use it as part of one of my first jobs using Houdini and it was a serious head-ache.

I do still use it, but I'm extremely weary of doing anything slightly complex in it, which is quite a let down as it has insane potential.

CHOPs, definitely needs more educational content out there, I've found a few, but they really only cover the literal basics.

I would vote for a houdinukestudio

COPS stands for Crashes Or Painfully Slow