The CGI Process

I have been thinking recently on the peculiarities of CGI and the challenges I often face in working with clients who have little or no knowledge of 3D animation. This complicates an already complicated process by adding a steep learning curve for the client and a whole ‘educational’ aspect of the project for me. Of course, clients don’t need to know the details of how everything works, that’s my job, but there are certain concepts that really need to be understood for a project to run smoothly.

With this in mind, I decided to write a guide for people who have never done this sort of thing before, much of it is common sense but I’ll cover some basic technical aspects as well; hopefully this will make the whole process a lot less daunting.

Don’t bluff.
If you’ve never done anything like this before then lay your cards on the table and say so from the outset. It will save a lot of time and confusion as the artist you’re working with won’t be making any misguided assumptions as to your level of knowledge. It will also avoid any complications that could arise from you being expected to understand certain terminology or be aware of any lead times that may be involved.

Read the Contract.
I have a standard contract in place that all my clients agree to on a job by job basis, it includes some basic terms and conditions and intellectual property rights and so on. It’s there for the client’s benefit as well as mine and is intended to protect both parties and ensure no-one is unclear about how the process will work. It’s not huge by any stretch of the imagination, just a couple of pages that you can get through in about 10 minutes and it must be agreed to before I start work on a project. Any artist you hire will have something similar and remember that this is a legally binding document, so make sure you read it properly.

Keep your secrets.
Whilst the majority of work I do is intended for promotional and marketing purposes (and therefore, by definition, in the public domain) there are still instances where confidential information is shared and, occasionally, where the final product itself is confidential. Whatever the situation may be it is imperative that you let the artist you’re working with know if anything is confidential from the outset. Have an NDA (Non-Disclosure Agreement) in place, get it signed and then both parties will know exactly where they stand.

Stick to the script.
The script should be the starting point for any animation project; yes you can work out a basic storyboard without a final script but, for the process to work smoothly, you need a locked-down script in place. The script will dictate scene length, key events to time animation to, particular sections that will need extra detail (as well as sections that won’t) and, of course, the overall animation length. This final point is critical because of the impact it has on working out how long the rendering process will be.

Know your responsibilities.
Whilst the artist you’re hiring will be responsible for driving most of the project forwards, it will be you who is responsible for contributing the basic content and guiding the artist with regards to the details. We can have ideas about how things should look aesthetically but we really do need to be led by you when it comes to the technicalities – after all it’s your product!

Have a deadline.
Even if your project doesn’t have a set delivery date it is usually a good idea to set a deadline to aim for so that things stay on track. If this is the case though make sure you decide on this deadline with your artist as they will be able to advise realistic timescales for completion. What you don’t want to do is pick a deadline yourself and give the artist the impression it’s critical to hit that date. If it’s a tight deadline they may end up working out of hours or hiring extra help to get it done, and that will all come at an extra cost.

Understand ‘sign off’.
This is something that applies across all kinds of creative media but I often find that, for one reason or another, clients think that sign off is somehow more flexible in CGI than it is in traditional media like print. CGI is extremely versatile and you can change anything you like; but it isn’t instant, there is no ‘magic button’ and every step in the process takes a lot of work. This is why I always get sign off on every key stage of a project to make sure that I’m not working on something that’s going to change further down the line.

Making a change to something that has been signed off is sometimes unavoidable since projects do change as they progress; but these changes range in impact and anything that needs changing after it has been rendered will be a major upset to both budget and timescale.

Give good references.
Be sure to provide good reference material for any styles, looks or materials that you want to replicate as these will be invaluable to the artist in trying to give you the final product you want. It will also save a lot of time and shortcut any potential misunderstandings with the artist trying to replicate a particular image you have in your head.

Size matters.
The final size of the video is a very important consideration, not only in terms of composition, but also in terms of timing as the video size you choose will have an impact on rendering, post-production and the final edit. In basic terms, the larger the frame, the longer it takes to render. For example a full HD frame will take roughly 4-5 times longer to render than an SD frame; considering that there are 25 frames per second of animation you can see how that would soon add up over the course of a full render.

You should always plan for the largest size you are going to need, files can easily be scaled down with no loss of quality but if you upscale from a smaller video size the image will suffer, in that instance (depending on how much you needed to upscale) the entire project would need to be re-rendered, processed and edited which, as mentioned above, could be a lengthy process indeed.

You can find a list of the most common frame sizes below:

Standard Definition:
4×3 (720×576)
16×9 (1024×576)

High Definition:
720p (1280×720)
1080p (1920×1080)

Conclusion.
These are some of the common stumbling blocks and tricky technical details that clients new to this kind of process will face, obviously the artist you hire will guide you through it all more thoroughly but, in the midst of a project, it can be easy to forget to go over the ‘obvious’ (to us!) details.

If you have any comments or questions on this then please get in touch, I’d be happy to talk to you.

CT.

Medical Illustration

Over the past month or two I have been working more and more on CGI Illustration and animation projects in the science and medical industries. This has been incredibly interesting work and some great stuff has come out of it; only problem is that I can’t show any of it as it’s all top secret!

To get around this issue (and have something to show people when I say I do work of this nature) I have recently started working on some concept 3D Illustrations in my spare time and you can see the first couple – for Virology and Haemotology – in the portfolio section.

Full size versions of these images are available for purchase on iStockphoto and if you have any specific illustrations you need (top secret or otherwise) then get in touch.

CT

Autodesk & Scaleform

I saw the announcement recently (through Jamie Gwilliam’s excellent blog) that Autodesk have ‘announced their intention to acquire Scaleform’…interesting stuff given the prominent position of 3DS Max in the games industry.

If you have never heard of Scaleform before then you’re like me, first time I heard about it was last Thursday. Looking into it though it seems like, while I may not have heard of them, I certainly know their product (well, from an ‘end-user’ point of view at least) as they are a popular choice for games companies looking for UI software. The software is based on Flash so fits into a creative workflow very easily as the diagram below demonstrates:

This excerpt from their website probably explains better than I can:

“Allowing development teams to quickly and easily implement 3D hardware accelerated game content and interfaces, including menu UIs, HUDs, animated textures, mini games, and even full casual games, speeding up development workflow, and freeing up programmers and artists to focus on what’s most important: gameplay and design. New UI components and workflow enhancement tools, make starting a new project faster than ever.”

This announcement sparks some interesting thoughts as to how integrated the Scaleform technology could become; will it remain as a standalone piece of software or will it become part of Max in a similar way to NodeJoe (which became the Slate material editor). If it is then the potential uses could be amazing, the ability to build interactive flash content from within Max would be awesome and would have a much wider appeal than just for UIs.

I’ll be interested to see how this develops and, hopefully, get my hands on it in one of the future Max updates. It seems to work in a similar way to technology we have been using to create 3D content that can be placed in PPT presentations; the 3D elements are created in Max, then exported and viewed with another piece of middleware (Flash in Scaleform’s case and DirectX in the case of the system we use), the content is then interactive in exactly the same way as the UIs in the video – you can even add trigger points to them which enable additional animation events.

One further piece of pure speculation, sparked by the recent XBR webinars, is whether any of this tech could be put to use for the redesign of the Max UI? Now that would be interesting…although I’m not sure how happy I’d be if the clean simplicity of the Max UI was ditched in favour of a floating, semi-transparent 3D version!

Here are some links to the announcement pages if you would like to read more on this story:
Autodesk announcement
Autodesk
Scaleform

CT

Ambient Occlusion in 3DS Max 2011

One of the key methods to achieving great results in CGI is ambient occlusion (AO), it adds extra depth to a render and takes a nice image that one step further. There are a few different ways to go about adding AO to your scenes but which is the right way to do it? The quick answer is that there is no ‘right’ solution for every project, as with most things it is a case of finding out which one will work best in the circumstances. So with that in mind I thought I would write a bit about the methods I think are the best and when, in my opinion, you should use them.

Just before we go over these different methods though we should probably touch on what AO actually is. Essentially Ambient Occlusion is a crude global illumination solution which works out how much shading geometry should receive based on how ‘occluded’ (or blocked) it is by other geometry. I have often heard AO described as a ‘nook and cranny’ shader and it helps to think of it as such; basically wherever surfaces meet you will get some shading, if a surface doesn’t have anything around it then it will be pure white. The black and white image you end up with can then be composited together with an initial render as the example images below illustrate, note the difference in shading and clarity from the first pass to the final combined one.

Basic method:
3DS Max can generate an AO effect using any renderer but the one I prefer (and have by far the most experience with) is Mental Ray. The basic method of adding AO to a scene with Mental Ray is simple:

1: Create a mental ray material
2: Plug the ‘Ambient/Reflective Occlusion’ shader into the surface slot
3: In the settings for the ‘Ambient/Reflective Occlusion’ shader find the ‘Max distance’ option. This is critically important as it determines how far surfaces will ‘look’ for occlusion – the smaller the value the smaller your shadows will be, the larger the value the larger they will be. This setting will be different depending on your scene and what you think looks right but as a general rule-of-thumb I go with 5-10mm for close-up prod-viz shots, 90-100mm for arch-viz scenes and higher for cityscapes etc depending on camera distance.
4: Next find the samples setting. This controls the amount of noise in your shadows, so if your AO looks a bit splotchy you put a higher value in. 16 samples are fine for test renders but for production renders you will need to up it to 64
5: Finally I always change the falloff setting to 0.75 but that’s just a personal preference, you can leave it set to 1 if you like.

Once your AO shader is setup you need to apply it to your scene. The simplest way to do this is to use the ‘material override’ option which is under the ‘processing’ tab of the render setup window. Just drag your mental ray material to the material slot in the material override section and drop an instance in there; this will now automatically apply your AO material to everything in the scene, easy! You can apply the material by selecting everything in the scene and applying the mat to all, but it is usually better to do it with the material override as you get to keep all of your current material settings and can just switch the AO mat off when you’re done.

One other thing to remember is that if you are applying this to a scene you have already textured and lit (which is usually the case) then you will need to switch off/delete all the lights, turn off MR exposure control, remove any environment maps, set the environment to white and turn off final gather – the ambient occlusion calculation doesn’t require ANY lighting, it is using the geometry only to work out how much shading should be in your scene so having any of the above on will just screw it up.

Bells and whistles:
So we have covered how to do a ‘pure’ AO render with the method outlined above but there are some problems with that, yes it does give you great ‘connecting shadows’ but since everything is black and white there is no colour bleed from the materials, also since the previous AO option has no lighting solution (and is purely ambient) there are no indirect light bounces – how do we solve this?

Well firstly all of our scene materials need to be MR A&D mats – these can be quite daunting at first but once you get into using them and setting up your scenes with physically correct settings you will see how versatile they are. In each material you will need to scroll down the options to the ‘special effects’ group and in here you will find an ‘ambient occlusion’ setting, once this is turned on it will apply ambient occlusion to the material.

The settings are similar to the other AO option in that you have a ‘samples’ option to control quality and the ‘max distance’ setting; however, you then have a check box to ‘use colour from other materials (exact AO)’, this is the option that gives us shadows with bounced light so, as a general rule, I have this checked. The only other thing to check is that the ‘shadow colour’ is set to ‘global ambient light colour’ - this makes sure that the shadows generated will use the global settings from the ‘Environment and Effects’ tab rather than a custom setting per material.

NOTE: If you are working on a large scene with lots of materials in it is quite important to get your AO settings right in the first instance, clicking through hundreds of mats to tweak settings can make you lose the will to live! There is a very useful script available on Joe Gunn’s website called ‘Material Tweaker’ which lets you specify material sets (which work in the same way as selection sets) and then make global changes to them. It works with Standard, MR A&D and V-Ray mats and whilst you can only make general changes to them the MR A&D set does allow you to switch AO on or off and adjust the radius (the script uses different terminology to the actual materials, basically AO radius = max distance). You can’t adjust things like sampling quality unfortunately but in that case I would leave it at 16, do some renders to see where there are problems with noise and then adjust specific mats accordingly.

Once all your materials are setup in your scene you will get a very subtle ambient occlusion effect which, because it is part of a true final gather calculation, has indirect bounced light and colour bleeding in the shadows. But as technically accurate as this calculation is I tend to find that the end results look a bit washed out and weedy, yes the indirect bounces and colour bleeding look beautiful BUT they also take some of the shadow impact and edge definition out which, to my eyes, makes it look wrong – so what do we do?

Mix and match:
What I tend to do on almost every project (but not all of course – this is another thing where artistic license comes into the equation and you work out from project to project when it is relevant or not) is use both AO methods so that you have beautiful shadows with indirect bounces AND nice connecting shadows with well-defined edges.

If you know about rendering in passes then this will seem very basic, but what I do is setup my scene first with MR A&D materials (with occlusion switched on) and photometric lighting then render out the first pass. I then open another version of that scene and save it as an ‘AO’ version, setup my Mental Ray material with the ambient/reflective occlusion shader and plug it into the ‘material override’ section, delete all the lights, set the background to white, turn off FG and turn off exposure control. It may seem strange to do this in a separate file but I find it easier to keep things organised this way, I don’t like having to switch everything off and on all the time in the main file, keeping them separate means I don’t forget to turn things back on again or accidentally delete some lights. There are other reasons why having your AO version as a separate file is a good idea but we’ll go over those in the next section. Once this is all setup and you have your renders it’s time to stick them together so we switch to the compositing software. I use Photoshop for stills and After Effects for animation (although if I’m rendering Exr files I will always use After Effects as it deals with them better than Photoshop) but you can use whatever compositing software you’re comfortable with, the concept is the same no matter what you use:

1: In Photoshop open up your 2 files, the first pass and the occlusion pass
2: Copy the occlusion pass and paste it onto a new layer over the first pass
3: Change the layer setting for the occlusion pass to ‘multiply’ and adjust the opacity to suit your taste
4: Done! If you switch off the occlusion layer’s visibility you can see how much of an affect it has on the final image, it really adds definition and gives your shadows a great sense of depth.

The fiddly bits:
There are a couple of instances where the methodology used above doesn’t really work, these are also when having a separate file for the AO starts to makes a lot of sense so let’s briefly go over them:

1: Custom AO radii
The basic method of adding a global ambient occlusion shader to your entire scene is great for getting quick results but, as discussed earlier, depending on their size/detail objects sometimes need to have different ‘max distance’ values – a scene where every single object has the same setting just doesn’t look quite right. So the answer is to ditch the ‘material override’ option and apply custom AO settings to your objects. You don’t need to set them up individually of course, you can go with a few AO mats that have different distance values – one for fine detail and small objects, one for medium and one for large. Setting your AO up this way is more time consuming but it does give better results.

2: Reflective AO
In the ‘ambient/reflective occlusion’ shader there is a check box titled ‘reflective’, this can be very useful in enhancing the realism of your AO solution when it comes to highly reflective objects. Obviously a surface should not have a completely uniform amount of reflection, specularity maps help with this but the reflection occlusion shader can be used as well to get better results. Instead of sampling directly out from the surface normal as with standard AO, when you check the ‘reflective’ box the shader works in a different way and takes samples from the reflection direction of the surface normal. This calculation then focuses on areas that should have very little or no reflection, such as in tight corners etc, and will give you a map to use in compositing which reduces the reflection in these areas.

The images below are a good example of reflective occlusion, note the difference in specularity on the seat fabric material the occlusion map makes in the final image.

3: AO with masks
There are some instances where you have objects with transparency; the leaves on a tree model for example where you have a basic plane and the leaf shape is cut out with an alpha channel. Now obviously if you setup your AO pass as described above the alpha channel will be lost and your AO pass will render the full planes rather than the leaf shape, to get around this we need to, again, abandon the ‘material override’ method and go for a more custom setup. All you need to do is apply your basic AO to everything in the scene then, for the leaves (or whatever else it may be) do the following:

1: Create a blend material
2: Plug the AO material into slot 1
3: Create a standard material with opacity, specular level and glossiness set to 0
4: Take the alpha channel used in the original material and plug it into the mask channel of the blend material

That’s it, you should now get an AO render which has your transparency masks included.

4: AO with round corners
One of the neat features of the MR A&D materials is the ’round corners’ option, this lets you create an edge fillet effect without actually generating any extra geometry. It is useful for simple things like walls or the edges of a book…basically it can work quite well with any straight edge (although for most things I still prefer to actually model fillets into them). If you do have any of these in your scene then you will obviously need them to be in your occlusion pass as well, if you don’t you will have some strange looking edges that are both rounded and straight at the same time! Again we will need to ditch the ‘material override’ option and apply AO to the scene selectively. Once you have applied the standard AO shader to everything that doesn’t have the edge effect on do the following:

1: Create an MR A&D material with the colour set as pure white
2: Turn the reflectivity down to 0
3: In the special effects tab turn ’round corners’ on and set fillet radius as desired
4: Plug the ‘ambient/reflective occlusion’ map into the ‘additional colour map’ channel

That should now give you an AO render with round corners where required.

Conclusion:
I hope this post proves to be useful to people, I started writing it thinking that I would just do a basic summary of the AO methods I personally use but, as I wrote, it got more and more detailed. Even so the methods outlined above are by no means exhaustive and the ‘fiddly’ options I have listed, which will undoubtedly give you better results, are ones I don’t use that often – if anything I tend to find myself just doing the quick AO + MR A&D AO option as it gives good results with a minimum of fuss. The other options are good to know though for when you need to go that little bit further.

CT

Working with Slate

3DS Max 2011 has been around for about 6 months now and one of its features was the updated interface for the material editor – the Slate. This is a move towards a node-based system; something that a lot of other software (Maya, XSI, Blender, Shake, Fusion, Nuke etc) has been running for a while and a method that is generally considered to be an efficient and intuitive one…but is it?

Well the old material editor in Max had a couple of problems – the need to reset material slots if you went over 24 mats always annoyed me, as did the way you had to click through everything to get to the ‘deep’ settings on complex materials. One big thing it had in its favour though was familiarity; I had been using it for about 10 years so, despite its quirks, I knew what I was doing with it…and then along came Slate.

I have to say that I found it an uphill struggle at first, the very different look and feel to everything was quite off-putting and my first reaction was that it didn’t really add anything productive to your workflow. After a day or so of perseverance though I started to see the benefits and haven’t looked back since – and what makes it so good? Well as I said I found it a bit tricky to get to grips with at first (and a number of colleagues still can’t see the point of moving away from the old system) so I thought I’d go through some of the things I’ve picked up and some of the key features from my perspective:

Customised Layout.
The layout of the Slate when I first opened it wasn’t quite to my liking, but fortunately if you click onto any of the standard windows and drag them around you will see some highlighted positions, drag the selected window to the position you like and drop it there. A feature I really like is the option to add a custom material group which you can drag all of your most commonly used items into, this saves a lot of time as the choice from the standard drop down menus you start with can be a bit bewildering , having a custom set to choose from is far easier.

Work Area.
That first issue with the old material editor (now called the ‘compact material editor’) about running out of mat slots is now gone as the working space you have to place your materials on is huge. Using the new available space it is now possible to create hundreds of mats in the same place; although in terms of organisation that might get confusing. To combat the confusion with complex scenes you have the option of creating new workspaces to keep things organised, these just sit at the top as tabs for you to flick between.

Everything at a glance.
This addresses my previous comment on dealing with complex materials, the new node based view means you can see straight away what the setup of your material is. No clicking through channel upon channel of mats and sub-mats to find out what’s going on, everything is just there to see. Another benefit of this is that if you use the same map in multiple places you can just drag the wire out from it to multiple slots/materials at once – it’s basically the same as instancing but is a much neater way of doing it.

Extra goodies.
There are also a few other really useful little things in the Slate; the option to load all scene materials onto the work area at once is handy as are the node/child layout options (the option to re-sort everything vertically can be very useful if things start getting a bit complicated),  the navigator window and the search functionality built right into the Material Map Browser window. I really love the way matlibs work now though, you just click on the arrow to the left of the search bar and select ‘open material library’, select one and it opens above your custom set and you can drag mats onto the work area; when you’re done just right-click the lib and close it. You can also open scenes as matlibs by changing the ‘Files of type’ drop down in the Open window – this is very useful as you can get quick access to previously used mats without having to add them to a custom library.

Space Invader.
One thing with the Slate is that it takes up a lot of screen space, it is lovely to use with a dual monitor setup (it does really need an entire screen in my opinion), working on one screen is of course possible but it’s a bit cramped and I found myself having to constantly resize and move the window to see what was going on…the title of ‘compact’ for the old material editor is apt indeed!

Personally I think the Slate is a big improvement over the old material editor and, as my first experience of a node-based system, I have indeed found it to be efficient and intuitive. I would say it is a great addition to Max, hopefully development will continue and new features will be added to improve the workflow of material creation even further.

CT

iRay: First Thoughts

The latest subscription centre update to 3DS Max 2011 included something I have been excited about since I first read about it over at the Mental Images website; iRay. It is, as their website states, “the world’s first interactive and physically correct, photorealistic rendering solution” and the important part of that sentence is the word ‘interactive’.

Because iRay is set up to leverage the power of CUDA enabled GPUs as well as the CPU power that rendering engines traditionally use it is capable of producing amazing renders in a fraction of the time usually required. But even if you don’t have the GPU power it still offers an amazing solution for visualising scenes interactively (it just does it slower) because unlike Mental Ray, VRay or any of the other traditional rendering engines it does not need to compute a complicated and time-consuming light pass before moving on to the final render, iRay starts rendering almost immediately and you can see fairly quickly whether you need to make adjustments or not. It doesn’t render in ‘buckets’ either so you see the whole scene rendering at once – it does this by rendering iteratively, so the image is very grainy and poorly defined at first but the longer you leave it the more refined it becomes, once you have an image you’re happy with you can stop the render and save it out – easy.

And it is that; easy I mean. Compared to the myriad settings of Mental Ray there is practically nothing to set up with iRay, you get 3 basic options to control how long it renders (you can set a time limit, an iteration limit or leave it at unlimited) and then some additional settings to control trace depth, image filtering, displacement and the ever-useful material override – that’s it! For someone who has used Mental Ray for so long the simplicity of iRay feels wrong at first.

The results can be spectacular though, whilst it certainly isn’t perfect (current problems include ‘fireflies’ (white pixels that don’t disappear from your renders) and lots of unsupported map types, including the new substance materials – although Ken Pimentel from Autodesk has said they are working on a fix for this) for a certain type of scene it looks like a great choice; take this simple prod-viz scene I put together as a test for example:

I decided to try and limit it by iteration so I could get an idea of where an acceptable level of quality was, I tried 1000 first but it was still pretty noisy so then went for 3000 which seems ok (although there are still a few specks of noise) – this took 49m41s rendering at 1024×683, not too bad considering my GPU doesn’t have that many CUDA cores to play with; also notice the caustics and DOF effects that iRay does ‘for free’. To highlight the DOF in this picture I rendered a section using the ‘blowup’ option.

So how did Mental Ray fare with the same scene? Well the render below took 27m44s with the following settings; image precision 1/16, FG set to draft with 5 FG bounces, Mitchell filter, Caustics off and the same MR DOF settings as for iRay. Whilst it is indeed quicker than iRay I think the DOF is noisier, the refraction and reflections aren’t as good and, even though I left them out on purpose, you really do miss those caustics. Of course all of that could be remedied, that’s the strength of being able to tweak everything with Mental Ray, and those changes would possibly bring the render time to about even with iRay. If I were to get another GPU then iRay would really come into its own, these tests on The Area were done on a machine with a Quadro 5000 and Tesla C2050 and while the render times and results are impressive so is the cost; a little over £4,000 for the pair, yikes!

I’m looking forward to using iRay on some projects but, as I said before, there are things that it just isn’t suited to – very large scenes for example (since the entire scene needs to be loaded onto the GPU) or scenes with proxy objects, I tried an arch-viz scene of mine that had hundreds of proxy trees in and it crashed pretty much instantly. I also think that whilst the DOF is a lovely touch it still doesn’t beat the flexibility of using a depth matte and doing it in After Effects with something like Frischluft Lenscare. These are small points though for what is surely a big step in the right direction; all we need now is Mental Ray to make use of GPU power and then that Tesla might seem like a good idea after all!

CT.

PS: For reference my workstation specs are – Dual Xeon E5520, 8GB RAM, Quadro FX1800 GPU.

Lets Get Linear

Linear Workflow (LWF) is a much mentioned topic in 3D circles which really seems to polarise opinions – some are all for it while others just don’t see the point. I have been using LWF for some time now and while I won’t try and write any lengthy explanations of it (as there are already many excellent articles online which I will link to later), I thought I would share personal experience on the subject in case it comes in useful for anyone.

So, I guess a good place to start is a brief explanation of LWF and what it means for 3D – basically almost all monitors ship with a standard gamma (luminance) setting of 2.2 whereas 3DS Max uses a default gamma setting of 1.0. To use a Linear Workflow you adjust the settings in Max to 2.2 and this gives you a gamma corrected display.

Why is this a good thing? Well rendering with a gamma of 1.0 will give you very dark results with lack of detail in the shadows, the problem is that most people (like me) will have been working in this way for years and think it looks right (like I did) and get around the problem of ‘things being too dark’ by either adding more lights to the scene or upping the intensity of the existing lights. So what’s wrong with that?

Well for a start more lights = longer render times as your rendering engine has to compute more bounces etc, and if you’re using area lights then this is an even bigger consideration. The other issue is when you’re using photometric lights and IES profiles they won’t be accurate and won’t produce the desired results – and why bother using physically accurate settings if you’re only going to ruin it by fudging the settings?

That’s what convinced me to change the way I work and embrace LWF, I was getting more and more into using physically accurate settings as much as possible so it was an obvious choice. However, I know a few other CGI artists who don’t care about physically accurate and just go for what ‘looks right’. Personally I think gamma 2.2 renders look right…and all without having to force your lights to compensate for your gamma curve.

If you want to learn more about Linear Workflow then have a look here:

My Mental Ray Community
CG Talk
CG POV

Is it real?

At a recent event I attended I was having a conversation about 3D visualisation with a group of people (some were relatively CGI-literate and others not at all), we were talking about the different approaches you can make on a project and the changes they will make both aesthetically and in terms of workflow. I briefly mentioned some of these approaches – like abstract or stylised, non-photorealistic rendering (NPR) and photorealistic – when one of the group said that ‘…3D will never be able to achieve photorealism’.

Initially I was quite surprised as, to be quite honest, I thought 3D visualisation was already capable of producing ‘photoreal’ results; just look at the subject of a previous post on this blog, or for that matter much of the product photography you see from major car companies (Audi, Ford, Land Rover, Mercedes etc), as well as countless other examples across commercial photography that you simply don’t notice because they are not visibly CGI. But then I started to think about the word ‘photorealism’ itself, what does that actually mean?

Of course we all know the general meaning; but in a world where photographs are modelled, styled, lit and shaped to perfection before going into Photoshop to be perfected even further…what exactly is real? Is the product shot that is actually photographed any more real than one produced using CGI? I think not. Ultimately the imagery we see is selling perfection, not reality, and photography and CGI are both perfect tools for generating that fantasy.

I realise that this long-winded ramble is so completely not the point the original guy was making, I just thought it was interesting that the term ‘photoreal’ doesn’t even really apply to photography anymore!

Perfect Imperfections

There are many similarities between traditional photography and 3D – the ‘rule of thirds’ or the principles of lighting scenes and framing subjects for example are exactly the same. But for all the similarities there are also differences – and there is one difference in particular that I find quite interesting.

That difference is imperfection and what photographers and 3D artists do with it. This is generalising a little but for most types of photography the photographer will be concerned with keeping imperfections out of their images, altogether if possible, as it will enhance their work. Conversely 3D artists try to add imperfections, albeit subtly, to their images for exactly the same reason – it enhances their work.

Now when I say ‘imperfection’ I’m not talking about the simple things like vignettes, lens flare or selective blur/depth of field as, in their place, they are useful effects for photographers and 3D artists alike. I’m talking about chromatic aberration, barrel distortion and noise. In photography these phenomena are caused by inferior lenses,  poor lighting or bad camera settings and require the purchase of some specialist kit to combat (Canon’s L and DO (diffractive optics) lenses for example) or extensive work in post.

So why do we try and recreate this in 3D? Simple, 3D is too perfect, too sterile and that in turn can make renders look fake; so to add realism we – through the purchase of some specialist kit and extensive work in post – add some subtle effects to simulate the physical distortion of real lenses and the noise or grain you get from film cameras. It may be subtle but it works.

This was all brought into focus (pardon the pun) for me recently as I uploaded some renders to a stock photography library only to have a couple rejected for ‘noise and lens distortion’. Their helpful tips on choosing different ISO speeds and adjusting aperture didn’t really help, I just removed the effects (making the images look less realistic to my eyes) and resubmitted!

The 3rd and the 7th

Architectural visualisation specialist Alex Roman took a years sabbatical to complete this stunning personal project – The Third and the Seventh.

Featuring some of the most inspirational examples of modern architecture in the world and, aside from a couple of bits of video added in post, made entirely in 3D we can’t stop staring at this in awe!

As the note beneath the video suggests, fullscreen it