I can't seem to find any learning resources for compositing 3D models
Sorry this it going to be a bit long.
For a long time I have been trying to find any tutorials or other learning resources including ones that I would have to pay for, that are showing a step by step workflow of compositing 3D models into camera footage. I am interested in a specific type of compositing where real life objects are expanded using 3D models. A good example that I can think of is actually used in a lot of movies, for example only a tiny part of a house is real, but it actually looks like a big mantion because 70% of the building is a 3D model.
I would really like to learn how to correctly light, texture and composit such models into real life footage, but all the tutorials that I find are really far away from being realistic in the slightest. I am able to find separate tutorials on modeling, texturing, lighting, but they do not apply as the methods used are for that particular scene only.
I can't find a tutorial that shows how to light a 3D model so that the light matches a scene, I can't find a tutorial that shows how to import 3D models into the footage and how to track them properly. I can't find any info whether these scenes are completed in the 3D software or in something like after effects. There seems to be a lot of back and forth between a compositing software and a 3D software, I can't seem to find any tutorials on what do you do in each one and the workflow between them.
I am able to create such compositions with photoshop because I find it easy to manipulate still images, there are brushes that make it super easy to composit objects into scenes realistically and I just find it easier to manipulate layers accordingly. With after effects and moving images in general I find it hard to import 3D models, if I do manage to import them I am lost as to how to make sure the object is lit correctly, I was never able to create working shadow catchers, tracking is always a pain and objects are always sliding, I also find it hard to manipulate 3D objects in aaf.
My modeling, lighting, texturing skills are weak, but I have a friend who does 3D visualisations and modeling using cinema4D with vray for a living and his work is looking really good, I wanted to collaborate with him to create some cgi scenes like the ones I described, but it doesn't matter that he is able to create realistic looking models, because I can't find anywhere how to implement his models into footage. If he lights a scene, he lights it in cinema4d for a specific interior visualisation, that obviously doesn't work for any random scene, if I import the models I import them without his lighting, when he renders the model with vray it looks super realistic, but then that doesn't work if I import it into after effects, does that mean that he needs to composit the entire scene in cinema4d, track it in cinema4d, light it and texture it in cinema4d? if yes, then what's the point of people using aaf to composit 3D objects going back and forth between the two rather than just using cinema or another 3d software?
I have so many questions and just can't find any answers, I was looking for courses, but didn't find any that would teach me this part of vfx.
Here is a good example of what I mean:
Especially the parts at: 0.37, 3.17-3.24 and 3.32
Well that's a very complicated topic. Im not sure why you're having trouble finding tutorials for those things though. a quick google search of "putting 3d model in scene with c4d and ae" returns almost 2 million results. The first few tutorials on that search page show exactly how to track footage, export camera to c4d and bring the model back into the AE.
The main benefit to using a compositing program like AE instead of doing it all in 3d is the ability to make quick changes. Instead of dealing with the longer render times from a 3d program. Essentially they should render out a model with multiple passes like diffuse, specular, ao, object buffer etc. Then you can combine those passes as layers in AE and tweak those layers to get the right look.
To implement your friend's models just open the c4d scene in ae and use the cineware effect to manage it with in the comp. Or better yet have him send you rendered passes of the model. He will need the camera tracking data from your AE comp and he can use that to match the camera on his end in C4D. Matching lighting is tricky. You'll either to match it in Cinema by using an HDRI that you've generated of the scene, or by carefully matching it by eye using lights, or using the skybox/daylight type of option.
My advice would be practice with simple models and scenes until you've gotten the basics like tracking/relighting/shadow catchers down pat. Trying to do an extension turning a house into a mansion is going to cause you to run into many additional problems that will be harder to solve. Im not sure how hard you've looked if you were unable to find useful tutorials.
Here are some links you might want to check out. You just need to practice the techniques and apply them on your own no tutorial is going to show you exactly what you need usually.
Thanks for the answer!
Maybe just my expectations are too high, I have seen some of the tutorials that you linked before and did not use them as I don't think they will help me with what I want to learn, all of those work well for some home made footage, they do not look realistic , they do not show specifically what I would like to do and so I didn't use them.
The first link is a still image so I might as well do that in photoshop, the second link works with a 3D model as a base and just shows how to track a 2D object which I can do, rather than showing how to implement that into real footage. The third link completely does not look realistic, it also didn't require using a 3D model because the object was so far away that a simple matte would do the job. The fourth link he just composites it in illustrator and photoshop and then pretty much just imports it into after effects.
I don't actually think that anyone working with million dollar movies uses cinema4D, I keep on hearing about houdini and maya, but that's the same thing with looking for tutorials for these two. I can easily find how to simulate a destruction of an object in houdini, but no tutorial shows a realistic composition of that object into the footage.
The only time that I can find something that looks realistic is in showreels, but then any tutorials are a massive downgrade. I was hoping that if such work is showcased that there will be at least 1 course that goes through an entire process of creating these super realistic scenes for big name movies. That way I would be able to learn the appropriate software, the pipline between what is used and so on.
The key is to apply the basic techniques to complicated projects. The people that make vfx on hollywood blockbusters work in large teams of highly experienced artists with massive hardware available. There's not a single tutorial that's going to explain how to do that. I don't think i linked anything that's a still image. Tutorials show you the basic idea that's why they're simplified versions. Instead of trying to find the perfect tutorial to show exactly what you want, you need to learn the basics and apply them yourself.
What's your skill level with compositing in general? Because million dollar blockbusters are really the pinnacle of the industry. The traffic light tutorial is essentially what you need. You said it looks completely fake but he's showing the PROCESS of how to do it. By learning the process you'll be able to try it in maya and when you hit a roadblock you'll be able to search for a specific answer which should yield better results.
If you're more interested in taking a structured course that might show you the more advanced stuff check out https://www.gnomon.edu/
Actually C4D is used in big movies. Anything that will fit the bill is employed. I've got a shot in here right now that uses Houdini to corral particles, C4D renders them and renders and animates some procedural elements, Maya does the animation and modelling, Maxwell is doing some rendering (out of C4D), Nuke is doing some of the composite elements, AE is doing some, Fusion is doing some... heck even Blender is being used for some smoke elements. Then it all goes for grading. Then it comes back and then it goes again.
The problem with big shops/movies is that all the jobs get compartmentalized. One guy does lighting, one gal does textures, one does roto, one does keying, ten build the models etc. There's no reason you can't learn all these skills, but... its a lot of skills.
Keep in mind too some of those shots in the example you gave have taken a team of dozens months to make less than 10 seconds of film. I'm not trying to dissuade you, I just want you to realize the size of cliff you are climbing. But climb boy climb! There's a heck of a view at the top.
Things that make shots look real:
-Atmospherics - haze, smoke, sun flares ( I don't mean lens flares), aerial diffusion, horizon color, shadow color (from sky color and other objects)
-Color - black levels are key, matching the shot color. Natural things are much drabber than you think.
-Lens distortion - lens length. A sure fire sign of CG is perfect lenses. No such thing in the real world. And CG guys tend to use too short of a lens for shots. What would a DP use? Get to know how film size relates to aperture size relates to lens size (and length) relates to chip size relates to 3D camera attributes.
-Exposure - what would have gone into shadow/got blown out, if shot on film
-HDRI environment/lighting maps
-High frequency vs low frequency noise: in foreground and background objects (respectively)
-Good composition to lead the eye around the scene, and distract from possible sins.
-Using elements shot at or created at the correct size. (which isn't always the final size they are in frame)
-Big comp sizes and high color depths. (otherwise there aren't enough pixels to resolve all that detail or not enough colors to go round and you get clipping or banding - reflective highlights are very very very bright and there's usually no true black)
Yeah, there are so many to choose from! My biggest struggle is indeed the amount of things to learn because as you mentioned it is usually one person focusing on a particular process thus making it the best it can be. I thought that the specific compositing that I mentioned would be not too much for 1 person to learn so I'm going to carry on doing so, but a long road awaits.
Thanks for the answer, I agree that it shows the process and I'm just being impatient trying to learn from the wrong side of things. Gnomon is something that I was looking for, but didn't find before! Seems like a comprehensive course for generalists for a comprehensive price.
It's hard to specify my skill level with compositing haha, I have a university degree in video production that included vfx modules using aaf and covered compositing so I can work with 2D compositing, but never had the chance to learn 3D.
There's also this recent thread in which I vented my spleen on a similar issue:
Its a shame that things have gone non generalist (or as we like to call it "multi-specialist") but in order to pump the work out at the level its at and in a timely fashion, it was a necessary evil.
One thing to keep in mind. Focus on what you like to do. Put stuff in your reel you like to do. Once you get good at something, that's about all you will do. We had a gal who was a rotoscope genius. She could pull a stable matte off the blurriest mess. She was her own worst enemy - she was so good, that's all she did. She ended up hating it. We had to make sure we rotated her around to different tasks and only called her in on the impossible shots.
One thing employers don't seem to get is that if someone is really really good at something, they are probably going to be really really good at something else too. It's not that they have a talent for that particular job, its usually more how they approach anything they do.