Large Patch count Rendering in Layers

[Shelton]

Ok this will show my limited knowledge of AM and animating with AM.

 

Some of the models for the short I am working on have 2500 patches all the way to 5500. There will be up to 7 characters in a scene at a time. There will also be a city scene with several buildings most of the items will be props as it were but still I can see a large patch count in the chor. So my questions are as follows.

 

In other programs you can render layers at a time. Does AM have that capablility? For example, as the girls walks down the street she will only interact with the sidewalk and street with her shadows. can she be rendered from the chor and then other elements rendered seperately? I have not tried it yet but I guess I could render the set/scene first with all the characters inactive. Then rendered each segment of the chor as active that I need and turn off the already rendered items. Then I could take each one of the renderings and add them into a video editing program such as Premiere, Vegas Video, etc and the render the movie from there.

 

Is this a way to by pass the large patch counts? I was looking at the TWO files and a large number the renderings I worked on, the file set up was already completed by Steve Sappington.

 

Any thoughts would be helpful on how to set up the scenes, chor

 

Steve

[robcat2075]

In other programs you can render layers at a time. Does AM have that capablility?

 

A:M doesn't have a way to permanently tag a model as being part of a certain group (AKA "Layer") for rendering purposes, but you can manually turn models off (Properties>Active ON/OFF )as needed if you want to render items separately.

 

And there is a way to get just a rendering of shadows.

 

 

I did this while testing some lighting for a TWO scene that had many characters and much scenery

 

 

All of this presumes you are somewhat knowledgeable about compositing issues and techniques.

[HomeSlice]

What Robcat said.

 

Do not have any lights in your set model(s). Do all the lighting either in the chor or in a separate action.

If you do the lighting in a separate action, make an empty model, name it "dummy_set" or "lighting" or something like that, drag it into the chor and drop the lighting action onto that.

Split the set up into foreground elements and background elements and make the foreground/background groups separate models. The ground would be a background element.

Turn off the characters and foreground and render the background without alpha.

Turn on the background, set it to "receive shadows" and "Shadow Only". Turn on the characters. And render with Alpha Buffer ON.

Turn on the foreground and turn off everything but the lights and render the foreground with Alpha Buffer ON.

Composite them together in either a new project in AM or in another editing app.

[Shelton]

Thanks guys excellent help. I have worked with several compositing programs for years and will be using them to complete the short. All the stuff I have done in the past has been one or two characters with very limited backgrounds.

 

Thanks so much as this has been a big concern while I was on vacation. I found myself thinking of how to do this or that and laying our the shot, I probably was not much fun to be around.

 

Steve

[heyvern]

I plan to try something like this with my Terminator spoof to speed up render times. Since I plan to use an environment map for reflections it should work for the shiny stuff. I figure the shiny robots will require the most rendering time so if I can render them separately from the background elements it will speed things up a bit. The "background" elements will be animated but not as much.

 

Don't many of the "big" studios (Pixar for example) use this technique? Rendering different elements in separate passes to composite? Is this done for rendering speed or other purposes?

 

-vern

[robcat2075]

BTW, this is slightly OT but with OpenEXR rendering you can actually get each light separated out and adjust lighting after you've rendered.

 

Don't many of the "big" studios (Pixar for example) use this technique? Rendering different elements in separate passes to composite? Is this done for rendering speed or other purposes?

Yes, I've read that they might render a background once and just slap it on an A:M style "Layer" behind a character in the foreground. Also some of their backgrounds are plain 2D paintings.

[heyvern]

In my case there will be shots of stationary characters talking at a table facing each other, the camera locked. Behind one character will be a large set of windows overlooking a large production assembly line. I figure the "window" and frame of the window can be a non moving single image render. The background elements seen through the window can be a short looping animated cycle. The foreground elements and characters would be another layer in the composite. The character with no moving background elements is even easier. If I cut back and forth and keep the wide shots to a minimum this should save a lot of rendering time.

 

-vern

[Fuchur]

Yep it will... I am quite sure that this is one of the features which wouldn't be too difficult to implement but which would have been against the notion of rendering all in one program.

(which isn't true because you could first render it and than compose everything together in A:M with Layers, etc.

 

In the end you can do it by yourself, and many are doing it like that. I recommend this method too, because it can really save you rendertime as Vern pointed out.

*Fuchur*

[Shelton]

Thank you guys for all the comments. Wealth of knowledge here. I am still a long ways away from that point. Just wanted to make sure I was not killing my efforts.

 

Steve

[Rodney]

I am quite sure that this is one of the features which wouldn't be too difficult to implement

 

The programming of such a feature might be fairly straightforward but implementing it in such a way as to satisfy everyone would be considerably harder. Rendering in Layers is but one of many features that fit into the general compositing workflow.

 

A relatively direct way of implementing this 'Rendering in Layers' yourself would be to organize your objects in folders on your harddrive in such a way that they can be replaced with 'empty' models. The folder structures would be otherwise exactly the same.

 

Example:

You desire to render only the Actors in your scene.

All models in your Actors folder are copies of your original Actors.

All Backgrounds and Props are empty models.

If you desire to render out Backgrounds without Actors you'd copy the original Backgrounds but empty Actors to the folders instead.

 

This is the example of how it can be done working with the resources behind the scenes.

Its not the optimum but demonstrates how to get from point A to point B.

If you know how to code a program you could create the thing.

 

You could probably do this even quicker just by creating a duplicate of your Choreography for each Layer you want to render.

 

Example:

You want to render only Actors in your scene.

Then you want to render only the Background.

Save your Choreography as 'MasterChor.cho'

Inactivate or Delete everything but the Actors in the Scene. Save the Choreography as 'Character Pass'.

Open MasterChor.cho again.

Inactivate or Delete everything but the Background objects in the Scene. Save the Chor as 'Background Pass'

Do the same for other layers as needed. "Overlay Pass", "Special Effects Pass" etc.

 

Finally, create another Choreography to combine/composite the whole thing.

 

Ultimately this last method will not be much more difficult than if you had to organize the Layers in a new feature. The important thing (and much more difficult if not impossible to program) is knowing what you need.

[Shelton]

Thanks Rodney

 

This was one of the ways I was thinking about. I had invisioned making a copy of the chor and deleting what I don't need. All the machines in the farm have plenty of memory and speed but still I was test rendering two characters plus a prop and AM crashed saying out of memory. So I knew it was time to think how I was going to render the short. I will keep you up to date.

 

Steve

[Fuchur]

We already have a feature which gives us a good startingpoint: Lightbuffers.

It is quite easy to use them... you just have to drag and drop lights into lightbuffer-folders.

The same done with any object (so lights, models, etc.) and the interface itself is done.

Of course there will be some stuff that we cant see yet, but shouting down any feature with "can be done by hand" or "not everybody will be happy with the inital release" will not help neigther.

 

I wrote a featurerequest on it and if Steffen want to implement it sometime in the future, he may. If not I understand that too... I just write the report to give some notions what could be implemented.

He is the one who chooses which he likes to implement. Of course that will take some time, but if we get the feature in 4 years or something I am still happy.

 

Till than, do it like Rodney suggested. Good startingpoint. I do it this way till now. Use a starting-chor and than just delete the stuff that wouldnt be in the same "level / layer". Embed and rename it and import the chor again. Do the next layer in the same way.

 

Here is another one: If you separate the shadow from the objects by rendering it in an alpha-channel, you dont need to use any soft-shadows. You can easily apply a blur-filter on it later in you Postprogramm or in A:M and gain a very equal effect (not totally equal, but most will not notice that the shadow is blured the same in any distance for most situations.

 

*Fuchur*

[Rodney]

It should be noted that there are other ways to get this 'rendering in layers' done as well. The two methods I mentioned are only examples.

 

Have you ever found yourself working with a feature and thought... "this is cool... if only it would..."

This is the nature of the beast.

 

Features are great and the more of them the better but one thing to keep in mind is that features are generally written to be used a particular way. When you create your own methods you may find yourself with a more flexible and useful means.

 

One of the greatest assets of A:M is that we can create our own features. The processes that work well with the project you are working on right now are the most important thing.

 

One of the problems with Rendering Objects in Layers is that stopping at the Object level is not going to be enough. Consider a simple situation:

 

You want to be able to composite different heads with a standard body in a scene (perhaps the body is exactly what you want but you know the head will need to be refined/changed). Its easy enough to render out the heads and body separately but when you go to composite it becomes a more complicated thing. If you didn't set up the renders correctly you might have unneeded elements composited over the top or showing through where not needed.

 

Each situation will be different but in the one I'm thinking of the solution would be to hide the back of the shirt for a rendered pass. We'd then place this new layer without the back of the shirt over the top of the 'back of the shirt' layer with the head/neck composited in between. We've then arrived at a solution for this particular scene and the head can be changed as needed.

 

Another method would be to use sequential decals to organize and layer the scene. Setting up a scene in this way would allow for more flexibilty than just rendering out separate elements in a scene. With this you could render all characters red, all props gray and all background elements light or dark green. Special effects and ghosted layers without the backs of models showing through them can be added to a scene. (Steps to reproduce this method we'll save for another day)

 

I guess what I'm trying to say is that much of this doesn't need to be programmed or confined to wishful thinking. We've already got these features.