Applying Special FX to the Camera

[brkn]

For simple things like that text, this refraction technique would work fine, but for anything that's much more complex, it's a bad idea.

Firstly it means that every pixel you see is going to be raytraced, which will slow things down considerably. Because of this, Full OSA (a side effect of raytracing) will be on for all your materials which will make them a lot slower to render, whether it's needed or not.

Secondly it forces you to make every material traceable so it can be seen through refraction, which prohibits you from optimising render times by excluding particular objects (usually large, or complex ones) from raytracing.

Thirdly, if you have any other raytraced refracting or reflecting materials in the scene, you'll have to raise the ray depth of the 'enclosing' object substantially in order to get decent results. This will also wreak havoc with render times.

These sorts of effects like the one demonstrated can be easily done in post, and that's how they should really be accomplished. Although Blender's compositor doesn't really have anything like it at the present, if we eventually get something like Shake's displace node, it should be really easy.

Anyway I'd be cautious advising newbies that this is a good generalised technique for all stills and animations, because it's not, and could get them into really bad habits ("blender sucks! why is this simple render taking so long!"). Simple scenes, yes, but beware...

[kernond]

Thanks for the crits!

This is an example of one effect that's possible using the concept of placing objects around the camera and animating them. The possibilities of this are huge when used creatively. So, please don't discourage others by throwing away the entire concept because the render times aren't totally optimized for this ONE particular effect! There are other materials that can be produced in Blender besides transparent/refracting surfaces. Also, have you actually tried using the technique or, is this strictly theory?

It's never a bad habit to try and be creative. Think more outside-the-box on this one.

Loved your work in Elephants Dream, BTW!

:)

[brkn]

Hiya,

It's good to see you sharing your ideas with others, too. It takes a fair bit of dedication to run a blog like that, so well done. I really do think that you should amend this article to at least mention that it will cause a slowdown, and touch on why.

Anyway if I can get on the soap box for a while, I'm not throwing the baby out with the bathwater, perhaps there could be certain effects that can be done this way that could never be done any other way and that can't be faked easily in post and are worth taking the speed hit for. That's fine. But a lot of the time that won't be the case, and I think it's important to mention that, and mention why, so that people can understand the issues well and make their own informed decisions. It worries me to see newbies get taken down a path where they think efficiency isn't an issue, or rather don't understand what causes inefficiency, because it can really hamper their progress later. While render times aren't really a problem when you're doing simple stuff starting out, they really can be as things get more complex and it's absolutely essential that one has some knowledge of what slows things down and how to optimise your scene. Heck, big studios even have entire departments dedicated to it (I remember seeing Cars had about 12 'optimisation artists' in the credits), it's very important.

I could give numerous examples of us having this problem in ED. In the 'switchboard bridge scene' long shot, when all the models came through to me to light and comp, the shot played back in the view at about 1/5 fps and wouldn't even render, it had more than 10 million faces and just crashed Blender on rendering. This was right at the start of the project, and we weren't used to the scale and complexity that would be involved - the switch plugs were modelled in a very 'freeform' way, with intricate detail and high level subsurf that would never be really noticed from a distance. I spent a couple of days going through all the models, removing all the subsurf that wouldn't be noticed, removing doubles, removing detail that wouldn't be seen anyway from so far away, optimising the mesh topology to use less vertices, etc. etc. I got the scene down to about 5 million faces, then it was actually workable. There were other different examples too, we had to stop using raytracing for Proog's reflection at the start since it was taking over 2 hours per frame to render. This was getting way out of control, because you've got to remember render time isn't only time waiting for your final animation to pop out, it's also test render time, time spent waiting in front of a monitor for your image to appear to see if it's right or not. We ended up using environment maps, which only took 20 or so mins per frame, a huge improvement. As a side note I always find it sad when I see new users boasting in the forums about how hardcore they are by having such long render times, maybe after they read that a pixar movie has some frames that take hours and hours. They should really be boasting about how fast they can render it :)

Anyway, if we hadn't known what was causing the slowdowns and had clues on how to go about improving them, we would have been pretty screwed. I know I'm ranting on and on, but render time really does matter and it's very important to know how to optimise (and that includes knowing when to fake the same effect in post). It may not be something that most newbies will immediately care about when they're starting out making simple scenes, but they need to know these things lest they be stuck at a newbie level forever.

cheers :)

(PS. yes I'm a bit of a hypocrite since in the time taken to write this post I could have started a nice wiki 'scene optimisation' doc.. maybe in the future, it's really needed!)

[kernond]

I really do think that you should amend this article to at least mention that it will cause a slowdown, and touch on why.

Yeah, I think that's reasonable, although, I didn't have any problems with the render times when applying this to other scenes that were animated along with camera movements. But, I guess anything can get a little out of control if you don't watch it.

I actually got the idea from the ED teaser (title effect) and the beginning of the actual movie (fish scene). I was also inspired by one of the recent demo games created in Blender for the physics contest. The game was the one with WWII dogfighting, done in B&W. There's a film grain effect that was created using animated textures (done in GIMP, or similar). The animated texture was on a plane object parented to the camera.

Anyway, I appreciate your expertise and will try to keep things like that in mind for the future. Again, I would like to say that this technique has great potential for things that otherwise would take considerable expertise and time. You have to, like you said, be mindful of the type of shaders involved.

(PS. yes I'm a bit of a hypocrite since in the time taken to write this post I could have started a nice wiki 'scene optimisation' doc.. maybe in the future, it's really needed!)

LOL, I was thinking the same thing when I got to that in your post. Seriously, a Scene Optimization guide would quickly become a MUST-READ for all blenderheads. Shame on you if you don't do it! :) ...you know, it's only one sentence at a time and, eventually, it's done. Considering your recent experience with ED, you (and possibly others on the team) are the prime candidate for the job. :applause:

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.