Sorry for the two threads but......

Is it possible to make an avengers movie with blender I mean graphics wise like the cgi characters texturing rendering and graphics.

Short answer: no. Long answer: nooooo.

Why do you say no? What exactly would you not be able to do in with Blender?

But seriously, like I said in the other thread, Blender is not the issue. It’s the talent, time and money that you’ll need to complete the work. If you had team of amazingly talented artists, a huge render farm, a good director with a clear vision and a few years, you could make anything. And you could probably use just about any major 3D program. Max, Maya, Modo, C4D, Blender, et al.

There really isn’t anything that the other programs can do that Blender can’t. If we are just talking about spec wise and the major features you would need to make an Avengers movie, I feel like Blender has it’s bases bases covered:

• All of the programs can Model characters right? Some even believe Blender to be superior at that. If you’ve ever modeled anything in Maya or Max, you might agree with that. :wink: Not all programs have built in Sculpting and Blender’s is pretty unusual in it’s implementation of dynamic topology sculpting. Blender also has built in basic re-topology tools. And afterwords, you can sculpt in finer detail with multi-res sculpting.

• Most of the programs have some way of either importing texture maps from a paint app or have some facility to paint directly in the app or even on the model it’s self. Blender does all of this.

• All of the apps have the ability rig the characters using a skeleton and attach controllers to it so the animators can work without to much trouble. Blender can do all that and then some. Maya has a built in muscle and skin simulation system which might be considered an advantage of Blender. But, Blender does have the ability to use other techniques that can achieve the same results. Pose space corrective shape keys, muscle bones other well placed dynamics can do the trick in Blender.

• All the program have the ability to animate characters using keyframes and shape them with an F-curve editor. Some are better at this then others. I’ve animated characters in Blender, Maya, Max, Modo and C4D. to me, Maya is the best, Max and C4D are the worst. That puts Blender somewhere in the middle.

• All of the programs can do some form of simulation but only Maya and Blender come with built in liquid and Fluid dynamics systems. Blender’s are much easier to use in my opinion.

• All of the programs have a built in Renderer that can render just about anything you need. Okay this is only half true because actually if you want to do serious work on Maya or Max, you have to buy a separate renderer. But, RenderMan, Vray, Octane and Maxwell et al. are all available for just about any 3D program you want to use (including Blender) so the argument that Cycles isn’t up to snuff is invalid. Especially when you would have to buy one of these renderers for the other programs anyway. But… I actually disagree that Cycles wouldn’t be capable. It’s not really missing any kind of show stopping features. Most of the effects like spectral color space and dispersion wouldn’t even be used on an Avengers movie. In my opinion, the biggest drawback to using Cycles is extreme slowness of motion blur on deformed objects (meaning characters). But like I said earlier, if you have a big render farm you can just throw a lot of computers at it and in the end you’ll get the results you are looking for.

Did I miss anything?

The main thing that the other programs offer is ease of workflow and production proven functionality. when people say you can’t do good work Blender because it’s missing X or Y feature, this is mostly what they are talking about. It’s the little things that come out in team based production…

For instance, anything you do is echoed in the script editor in Maya. You can just watch the commands scroll by as you work. You can also just highlight this text and drag it onto the tool shelf and it makes a button for you. So, any kind of repetitive task can easily captured and re-used later. It’s also a really great way to store commonly used scripts and have them as simple button at your finger tip. Another example: Maya’s smooth tool never flattens out you model like other programs. It maintains the surface curvature and volume. This is really useful when modeling. And another example: Maya’s render layers are much more powerful than just about any other program in just the simple fact that anything in your scene can be overridden with an entirely different value for a specific render layer. The only other program that can do this is Modo.

But, you can also say that Blender is actually pretty well proven in production too. In fact, every open movie that they do advances the program by leaps and bounds. It also helps to refine the workflow with each movie they do. Advanced features like dual quaternion skinning, Pose space deformation, heat mapping and Bendy bones, came to Blender back in the days of Big Buck Bunny. These are features that Maya and Max are just now getting! And they don’t even have anything like Bendy bones. So, it’s all kind of relative when you think about it.

P.S. I’m leaving out XSI because it’s a dead program now folks.

Stability? Scaleablility? Quality?
Did anyone rendered an epic Hollywood like VFX scene with Billions of Polys, GI, Motion Blur, DoF, Volume graphics, physics, fluid dynamics and whatnot in a 4K resolution? If yes, i’d like to see it. As far as i know nobody has done it on that scale yet. Which is important.
It takes a lot talented people, huge amounts of money and time, a multitude of very expensive professional hardware and software to get such results.
I am not talking about one or two characters animated into a filmed scene, i am talking about the money shots: a couple of characters fighting in front of an flying aircraft carrier which is crashing into a modern city skyline.

Blenders Physics, Fluid dynamics, smoke sims seems mediocre and limited compared to Houdini, Realflow and FumeFX.
You can’t just throw more computers at the problem and hope it works.
If OP want to know if its possible to archive the amount of work and quality like one of the Avengers Movies with Blender alone then the answer is unknown, because nobody has ever done something like this.
I am willing to bet on a NO because like i said earlier i doubt that its possible to archive flexible photo realistic results for large scale physics, fluids and fire/smoke/explosions.

Yeah Indy’s right.

I’ve atvhed tears of steel but it doesn’t look too realistic any examples?

Tears of Steel is 4 years old. And made by a few people. many companies are working together on a film like Avangers to get these reults. If Blender can do it? Nobody knows until Hollywood tries

I would say no

because of the are actual limitations in Blender…

There is no muscle simulation that is based on something like finite elements or finite volumes so something like Hulk would be out of the question.
no proper displacement in cycles (currently wip)
motion blur in cycles is extremely expensive renderwise
Hair tools are limited
simulation tools are limited
no udim, no ptex
linking is limited (correct me if I am wrong but its currently not possible to link two characters that reference the same file in Blender)
viewport can’t handle high poly scenes
no render api so external render engines can be a PITA to use
limited shape key (blend shape rigging tools) I have yet to see someone create blend shapes for something with 50000+ polys for the head

and so on

@tyrant monkey, its is possible to link multiple characters using the same blend file
(as long as you don’t have circular references it should work fine).

The muscle system for avengers was written in-house, so likely you couldn’t use any off the shelf software, without developing your own tools ontop of it.

@ideasman42 interesting I remember someone saying that wasn’t possible never actually tested it out myself(just did)

What isn’t possible is linking the same character (or rigged group) from one file multiple times and applying different animations to each instance. That is a current limitation.

Sorry if it seems like I’m picking on you specifically but I just want to address these individually.

I really don’t think you Need a muscle system. As Campbell pointed out, a lot of those systems you see in movies are custom anyway. And, if you really found you couldn’t live without it, one of the nice things about Blender is that if you had the money to make a movie in Blender, you could probably hire someone to write that feature for you. Maybe even the Blender foundation. And then everyone would benefit from it in the end. :-

no proper displacement in cycles (currently wit)

It’s mostly in there now. Download the latest builedbot. If you started to make an Avengers movie in Blender today, by the time you were ready to render is would be done. :wink:

motion blur in cycles is extremely expensive renderwise

Yes, this is very true. Maybe some optimizations will come in the future. Again, the nice thing about Blender is that you can influence the development way more than you can with Autodesk.

Hair tools are limited

How so?

Actually, I disagree with this. I agree that Hair physics are limited but hair in general is one of the areas where Blender excels. The rendering is just beautiful. The grooming tools are really nice too. Have you ever used hair tools in other programs? I use Maya, Modo and Max on a daily basis. Yes, hair farm is more advanced than Blender’s hair tools but Blender’s tools will totally get the job done. You don’t need hair farm or Yeti to make good looking hair. They just break down the different things you can do with hair into a procedural workflow so that you can change things down the line. I’ve seen hair in Blender that looks amazing. If you had enough talent, you could get the exact same results as you would with the other tools.

simulation tools are limited

Again, how so?

What part of the simulation system are you talking about? Smoke? Liquids? Cloth? RBD? Dynamic paint doesn’t even exist in other programs. Smoke/fire looks good enough for me and it’s quite fast when compared to FumeFX and Maya’s fluid dynamics system is dead slow. I agree that there are more advanced tools to influence you sims in other programs like realflow or Bifröst. But I’m sure given enough time, money and talent, you could figure something out that looked pretty decent.

I agree if you are referring to the lack of a good fracture tools. I do know that a lot studios don’t actually use these when breaking buildings and smashing cars and stuff. They have a team of modelers build all the custom shapes and destroy with normal physics. The thing is that most fracture algorithms don’t actually look realistic or are of limited use because they are good for broken glass or shattered crystal. But I agree that at some point you will need to do this too. :wink: There are various fracture tools out there for Blender. Even a fracture modifier that’s been in development for what seems like years. :wink: So, I’d have to look into this more.

no udim, no ptex

Now you’re just being silly. These features are totally not needed to make an Avengers movie.

linking is limited (correct me if I am wrong but its currently not possible to link two characters that reference the same file in Blender)
Corrected. :wink:

viewport can’t handle high poly scenes

And they can in other programs? Folks, I’m beginning to think most people here have never used Maya or Max and they just think they are so much faster. The issue with Blender’s viewport is not the 3D view being able to handle millions of polys. There’s a whole sculpting engine built into Blender. It can handle a lot. The issue is being able to handle them in edit mode. Believe me, I use these programs all the time. They have the same issues and choke just as often. You have to use tricks and workarounds all the time. For instance (no pun intended) you use render time proxies anywhere you can. You break up meshes into smaller ones so you can work on them. You re-topo and bake detail down to displacement and normal maps. This is all common practice in larger studios.

no render api so external render engines can be a PITA to use

There is a render API it’s just that external render engines can’t access Blender’s memory directly. However, you are right that this is cumbersome for some renderers. But, I would not say that it’s a barrier to entry. The Vray for Blender implementation uses a custom build that renderers directly in the interface so there are work arounds.

limited shape key (blend shape rigging tools) I have yet to see someone create blend shapes for something with 50000+ polys for the head

I think you are imagining a workflow here that does’t actually exist. No one does this. although, I should point out that 50,000 polys aren’t that much and Blender could actually handle that but I think you probably just forgot a zero.

Seriously though, A lot of the screen shots and youtube videos you see where someone is manipulating a high poly face for a movie are actually just showing the sub-d mesh. High poly meshes are actually hard to work with not because of the viewport speed for moving those points around but because it’s difficult for the artists to sit there and have to tweak every point. If you look at “The art of moving points”, he breaks down exactly how much detail you need in a feature film model. the guy used to be the lead character modeler at Pixar. The “Original” Hulk model may have had hundreds of millions of polys but you can bet that the animators never saw those models. You can bet that most of that detail was baked down to displacement and normal maps and you can bet the the model the TD’s rigged was a much more agile and reasonable mesh to work with. It’s hell in any program to have to paint weight maps on a 3 hundred million poly model.

The other thing I want to point out is that Blender’s shape key tools are actually quite advanced. It’s actually one of the areas where Blender excels at. Blend shapes in Maya are actually super basic. There isn’t even a mirror blend shape tool built into Maya. Not only that but tools that Blender has had for many years now, like the ability to build shapes ‘in place’ on the model and pose space deformation were only just addd to Maya a few months ago with version 2016 ext 2.

The one area where Maya may have an advantage is really just a workflow thing: In maya you can use a Blendshape node like a mixer for different deformation systems. This is because all the models used for blend shapes stay ‘live’ even after you’ve added them to them main model. So you can actually have a few copies of the main mesh, hook them back up the original, rig one with joints, another with lattice, another with correctives, another with a muscle system, etc. And, they will continue to deform the main mesh in varying amounts based on the amount that you dial in that blend shape. But again, this is really just a workflow thing and not really necessary for feature film work.

All in all, my point is just that I don’t think it’s impossible to achieve convincing results with Blender. I think the tools are mostly there, it just takes time and talent. Would it be easy? Hell no. But nothing worth it ever is.

EDIT: Oh! I just thought of something that would be hard to do with Blender: Crowd simulation.

This isn’t really something you do in Maya either. We always have to make individual versions of repeated characters so the name spaces don’t get messed up. However, recently we’ve just been using Alembic point caches instead of linking fully rigged characters into our render scenes.

So can’t C4D
Yet it’s still used for movies.

We’ve made use of it for limited things, the problem comes when you duplicate a reference to a file that contains nested references. The namespace only increments for the first level. So if you have a rig file referencing a character model (actually it was a semitruck in my story, but w/e) and duplicate the reference to the rig file, shit hits the fan. If you just have a simple asset where everything is in one file, it’s fairly useable.

BlenderArtists forum, summarized

Right, I read this as multiple characters which shared a common blend file, You can link multiple different characters, that all link to the same material blend file for example.


No you really do need a muscle system if you want to make an avengers movie…whether those kind of tools are available in off-shelf software or in-house tools is irrelevant. Going by the OP’s question I assumed he meant the animation,rigging, modeling, rendering, animation all happening inside of Blender.

I can’t see how you could do a Hulk who is shirtless and muscular and get all the skin sliding and correct muscle contractions and relaxing without one. Maybe you could do it if you were going for 2003 Hulk but not 2012 Hulk. Those kind of characters have pretty solid muscle sims from the head down especially if the are key characters. If you used Blender with another tool maybe you could do it.

The 50000+ polys on the face is probably low by todays standard. Pipelines my differ but that is what the Jake head from Avatar had in 2009, and it is the raw polygon count…IIRC(it’s been a long time since I read about that movie) it’s because all the wrinkles, frowns etc are driven by geometry not displacement maps.

I suspect the Hulk and all the other digital doubles in that movie would have faces with similarly high polycount if you want them to hold up at 4K and look as realistic as possible.

Depends, is that with or without the teams of software developers and TD’s that worked on the movie?

TBH, not sure if these projects would be practical with any off-the-shelf software.
It’s common for studios to do in-house development with whatever tools they use.

This is a huge investment, so its likely they wouldn’t do it unless they needed to.

In fact, here there is an addon to build muscles: