5mins 55sec 36meg Quicktime H.264 MOV

Offset tracking?  What about tracking in general, Matt?  Hold your horses and steady on!  I freely admit that this is an old tutorial I did for the University course I taught, so apologies to my previous students . . . please don’t let this make you feel any less special.  You are all unique snowflakes!

Actually the real reason is that upon reflection my original tracking video might raise the ire of Teh Appel Corporation as I used the standard footage provided in the shake tutorial book . . . soooooo, rather than risk a multi-billionaire dollar lawsuit and put Steve Jobs already strained health under more stress, I will look at doing a re-edit with some techniques I learned on ‘Australia’.  

Now, these concepts can be applied to pretty much any compositing software.  As far as I know they all have the ability to offset or append a search region.  Why do we do this?  The main reason is because your main tracking points become obscured or head offscreen.  The thing is, as an aspiring compositor or 3D person, you might not know that this is possible, so you will be fighting your footage trying to get “the perfect track” on a point when it either just can’t be done or you can get a better track quicker and with less time wastage by choosing better tracking data at some point in your clip.

So, without further delay, check it, like, totally out.  As always, comments, questions and additions are welcome.  Only unnecessarily over the top profanity is frowned upon.

Offset tracking. Sexxy, no? No.

.

16mins 38secs, 54meg Quicktime mp4 mov

I finally got my desktop fired up and grabbed this video I made last year for class.  I have a 50meg version of this I will try and get up soon if 80meg is too big.  I have used Simon Reeves’ excellent breakdown on his website www.bankaffairs.com.  Check it out for some very clear examples of using RBG passes for compositing in CSI Please note that all images are copyright Simon Reeves and are to be used for education purposes only.  Check out Simon’s main site at www.simonreeves.com.

Note: for clarity (and my own workflow), I have reordered the channels of the RBG passes before feeding them into nodes.  You don’t have to do this in Shake.  In the node being masked you can choose which channel you want to use as the mask.

Images (c) Simon Reeves

This topic came up recently on CGTalk and I also covered it in the compositing course I taught at the Queensland College of Art.  I do have a video screencast as well that I did that I will try and dig up as I think it is on my main PC which is in pieces at the moment.

Inspiring, no?

The question was: “How do I use ID passes (or RGB mattes) in my compositing?”

I guess the first place to start is to remember that compositing is all about colour channels.  Pretty much the first thing you do when you get your plates is check the colour channels (red, green and blue) to see what kind of information is there.  When you are working with straight CG images, it is a bit different.  You will often have various black and white mattes rendered out for you to isolate various objects in a scene.

As we (hopefully) know by now, mattes are essentially greyscale (grayscale for you North American types) images that affect an area of your based on the intensity of the pixel colour value, usually with pure white affecting the change the most and black affecting the least.  These can be things like images, patterns, keys or rotoshapes.

The good thing about CG is you can output just about anything relating to your render.  Diffuse, specular, shadow, reflection, refraction, velocity, ID’s . . . the list is long.  The most basic pass you will get will be something like a foreground character that has it’s alpha channel already pre-multiplied (RGBA).  That is, you could stick it straight over a background as it already has an alpha channel.

Often, however, you will be given a pass that has an object, or several objects, shaded pure red, pure green and pure blue.  Why?  Well, we know that we can use this colour information to mask our objects.  By using or reordering/shuffling our channels, we can isolate a particular object using our RGB.  

Let’s look at an example scene.  Behold the power of Autodesk Maya.  I fully realise that this scene is beyond average and should not be used as an example of my actual 3D skills.  I am trying to make this as simple as possible.  No, really!

Basic colour pass

Now, our very nice lighting and rendering department has given us not one, but two extra passes to help finesse our scene.  The first is a straight RGB matte pass:

RGB matte pass

The second pass is an RGB lighting pass.  This is essentially our scene with 2 or 3 lights (key, fill and rim) and a basic shader.  Again, the lights are pure colours:

RGB lighting pass. Only red and green in this example.

So how do we use these passes?  I am going to use Shake and the reorder node for clarity here.  In Shake, and other compositing apps, you can often just choose in the node or effect which channel you want to use to affect that node or effect.  I am a big fan of clarity though, and as such, I prefer to use the Reorder node so I can instantly see which channel I am using:

Shake script overview

So, as you can see, I have split off three Reorder nodes from the tips_RGB_matte image:

RED

tips_RGB_matte RED channel

GREEN

tips_RGB_matte GREEN channel

BLUE

tips_RGB_matte BLUE channel

And here are the reordered red and green channels of our tips_RGB_lights pass:

RED

tips_RGB_lights RED channel

GREEN

tips_RGB_lights

ALRIGHTY!!!!!  We are pretty much ready now.  As we can see, now we have a bunch of nice clean mattes which we can use to modify not only the objects in the scene, but possibly use to relight the scene entirely.

Soooooo, going down my super mega awesome compu-tron 2000 script, we can see that I have brightened the cube, multiplied the sphere by a strong green and added a slightly horrific shade of orange to the cone.  Not brilliant but you get the idea.

Brightening the cube

Greening the sphere

Orange cones are good cones

On to the basic relighting, we can use the two lighting channels to change the colour and intensity of our key and fill lights.  To change the intensity I expanded the alpha after the reorders (essentially a levels call):

Relighting using the key light (red tinge)

Relighting using fill light (green tinge)

It’s possibly not as clear as day and night in the above shots, but as you can see, with just two extra passes from your 3D application, you suddenly have a lot more options for adjusting the looking of your comp without fussing about.

As always, if you have question, comments or suggestions, let me know below.

9mins 53secs, 24.6meg Quicktime 840×526 h.264 mov

Here is a cool tip that was passed on to me by one of our supervisors on Australia.  The idea is to use a colour lookup curve to shape how the edges of your matte are blurred, pushing the white pixels out to the edge or pulling them back into the core of your matte.

As always, if you have any questions or tips post them here. I would be especially happy if you have your own blog and can show how to replicate these techniques in other software. I just don’t have enough time or knowledge to do the tutorials in each application.

.

10mins 02secs, 21meg Quicktime h.264 mov

Alrighty then, that’s enough doom and gloom for a while!  Let’s get back to some tips, and what better way to start the week than with the almighty lightwrap.  Easy to overdo but a crucial part of bring your comps together, even if you are doing an all CG project.  I am sure that lightwrap in 3D would take a bugger of a long time to calculate and render, just like depth of field.

Our video this week will look at how to do a basic lightwrap in shake and explain why each step and node is used.  Yes Nuke has a lightwrap node.  Yes I am sure that AE has bazillion lightwrap plugins to do the job.  But you’re not just a monkey pushing buttons are ya?  I hope note because that means when I ask for a raise they are probably going to give the work to as you will happily work for grubs and the chance to groom the HR staff.  When a monkey does that it is cute.  When I do it it’s creepy.  Go figure . . . 

Click on the image below to get a full sized screengrab of the script so you can follow the tree flow.

11mins 14secs, 30meg Quicktime h.264 mov

Alrighty, due to overwhelming demand, here is a quick video giving you an overview of the mind-space of node-based compositing.  By “overwhelming” I mean “at least one person”, but as anyone who has worked in the media/advertising industry will tell you, every person who takes the effort to write in represents a vast segment of the population.  In this instance I assume it means that there are 4 trillion people reading this blog.  Sure I could actually check my statistics, but never underestimate the power of denial . . . 

In any event, hopefully this gives you an idea of how you think when dealing with nodes vs layers.  This video isn’t meant to go into detail about the composite, it is more just to show you the general flow of things and how they relate to each other.

As always, if you have any questions, post them here and I will do my best to answer them.  If I sound confused it is because I ate approximately 3 head of cattle at last weeks BBQ.  I apologise to any vegetarians reading, but I have it on good authority that the cows I ate were real jerks.

Wellity, wellity, wellity!  Now that’s the question, isn’t it?  Or to put it another way, it’s not really the question but it is a fairly common (and understanable) one.  Let’s make this clear:  I am definitely filing this post under the ‘Opinion’ tag

There are a lot of different opinions on this and of course you could easily say “It’s the artist, not the software”, which is about as helpful as being kicked in the family jewels.  The thing is, even if you aren’t a raw beginner, learning software takes time and effort and it makes sense to focus your time and effort on something that is going to give you the best bang for buck.

Take 3D apps for example.  XSI, Houdini and Blender are fantastic apps, although XSI took a hit recently IMO by dropping their cheap Foundation software.  However, the cold hard steely fact is that Maya is king.  Whether it is ‘best’ or you like using it or not, the fact is that it is used in more production facilities than other software.  Sure 3ds max has a solid foot in the door in gaming and dominates architectural visualisation and Houdini kills with VFX, the fact remains that if you are going to spend 3 years studying something, you want to study something that is most likely going to help you get a job.

So, back to compositing apps, which one?  Before I give a brief breakdown on each of the major applications, let’s differentiate between node-based and layer-based compositing.  You may have heard the terms before, with an aura of mystique and high-end awesomeness surrounding the fabled “node-based compositing”.  

So what is it?  Well, node based compositing is essentially like a flowchart.  You have a node for an image, if you want to blur it, you attach a blur node underneath it.  At the end, your composite looks like a ‘tree’ with several starting branches all converging in the final output node or nodes.

Layer-based applications deal with the comp like Photoshop.  You have an image or images and effects or operations stack up in layers on top of them.  To merge them you can do pre-comps (render out branches) or create groups and then merge them.

So why one over the other?  Well, in my (and many others opinions), the advantage of node-based compositing lies in the clarity of your compositions.  Very complicated comps can be followed very easily, because you just follow the flow chart.  Given it is a flow chart, at any part in a branch, you can branch off again, do some adjustments and then just come back into the main flow.  This allows you to experiment very quickly and efficiently.  This kind of thing can be done in layer-based compositing apps, but it can get very unwieldy for complicated shots, or shots involving a lot of elements.  

The flip-side of this, however, is that layer-based compositing apps tend much stronger in their animation/timeline capabilities.  Doing motion graphics in Shake is not my idea of a good time.  In fact I would rather stick a pencil in my ear, and a blunt one at that!

So, with that it mind, what app you want to use may come down to what industry you want to work in.  If you are really into design and motion graphics, you would be nuts not to focus on After Effects and maybe know a bit of Shake/Nuke/Fusion.  If you are wanting to go into feature film, it’s node-based all the way baby!  It is worth noting that both Nuke and Fusion both have full-featured free Personal Learning Editions, with Nuke now having the upper hand in free training materials to get you started.  Also, the last time I checked, you can use the student editions of the Adobe suite applications to do commercial work, which is a massive bonus as the discount is for students is pretty decent.  Of course, double check the Adobe website to see if that is still the case.

Shake: (Node Based)  Undisputed king.  As far as I know, still the most used compositing application in high-end facilities.  The price drop to $500 a few years ago put a lot more seats out there and as a freelancer, it is a great deal to pick up a copy on Mac (Unix pricing is still $$$$).  Shake is a great place to get started as it really is very straightforward.  What you see is what you get.  But, and this is a huge but, Shake is dead.  Stone cold dead.  Dead as a Dodo.  It is no longer be developed, no longer being supported and as per Apple’s usual mode of business, there is zero information on anything to replace it.

Nuke: (Node Based) The “new” kid on the block.  Nuke was developed as the inhouse compositing tool for Digital Domain for many years (although sold commercially) up until about version 4.8 or so.  The Foundry bought out Nuke and started pouring some time and actual development dollars into it.  Version 5.0 was a major upgrade, especially in terms of interface and it was clear that Nuke was looking to pick up Shake users, and it looks like they are being successful.  Given that it was developed in direct response to what their artists needed, the tools are very flexible and powerful and immediately make sense.  So much of the stuff I am learning in Nuke has me saying “Bloody hell, I wish Shake could have done that”.  Impressive.

Fusion: (Node Based).  Fusion has been around for a while and version 5.0 was a significant update.  I have only had a bit of experience with Fusion but it seems to have a solid pipeline and in-software 3D particles and 3D comping.  Seems to be popular with a growing number of smaller studios, however it is definitely capable of doing the big shots.

After Effects: (Layer Based).  The king of affordable desktop compositing apps and the reigning champ for plug-ins.  AE has definitely moved on in the last couple of years and is a very capable piece of software.  You can operate in 3D and with the plugins, you can get some excellent results.  Used mainly for motion graphics/broadcast design stuff, but is fully capable of doing high-end TV and feature film work.  Excels at time/animation dependent comps but intense film compositing can get complicated when dealing with many layers which need pre-comping etc.

Combustion: (Layer Based) See description of After Effects, but divide everything by 2  I love Combustion and started out on it, but it is no secret that it hasn’t really progressed over the last 4-5 years.  Also, don’t believe anyone who says that a benefit of Combustion is that once you learn it you will be able to jump on a Flame/Inferno suite.  Apart from the fact that some of the user interface looks kind of the same, they have nothing in common.

So, the winners are . . . BUP BADDA BAAAAAAHHHH (from Zoolander ) Nuke and After Effects.  Chances are that as a student you are probably going to be exposed to After Effects in your courses and that is not a bad thing, but I would also suggest getting the Nuke PLE and all the free tut’s just to see how it works.

Remember, it’s all about the principles.  If you understand the principles of compositing, then you can quickly apply them across different apps.  We are lucky in that compositing concepts transfer relatively painlessly across apps, as opposed to 3D where you often face very different ways of doing the most basic things.

‘Til next time.

Matt

Alrighty guys.  I have been spending a bit more time with Nuke, getting myself acquainted with it’s way of doing things.  This isn’t going to be a review, or even a “which app should I use” post (more on that in the future), but more my first impressions.

• Firstly, it’s great to see Nuke going ahead with a strong development team, something that obviously was not happening with Shake in the last few years and will not happen (Shake not being developed for, in case you didn’t know.

• Following on from this, with it’s sale to The Foundry, the Nuke team has made a lot of effort to make the software more friendly to artists who are moving from different packages, especially Shake.  I used the PLE (Personal Learning Edition) of Nuke back around 4.7 or so, and while it was obviously powerful, I felt the interface was completely utilitarian and not very obvious.

• Lots of on screen controls soon become very intuitive.  The ability to have more than 2 viewers, ala Shake, is a god-send.  The onscreen scale/rotate tools and proxies for objects when you translate them are excellent.

• The layout is completely customizable, letting you set up your workspace.

• Very logical way to set up proxies for hi-def and film-res elements.

• Am just getting into gizmos (essentially macros) and the use of python scripting is great.

• The use of channels instead of piping masks in everywhere is a great concept and one I am still getting my head around.  Seeing as it is one of the most powerful features of Nuke, I need to spend more time with this.

• Roto tools seem a bit clunky, but frankly Shake was nothing to shout about either.

So far I have only really scratched the surface, but the more I use Nuke, the more comfortable I will be with it.  It is still tempting to go back to what I know in Shake when I need to get something done quickly, but I am consciously trying to do more work in Nuke.

If you wanted to get started, The Foundry have a PLE version of Nuke for download and have a tonne of training materials, including a bunch of great introductory videos from the guys at FXPhD.