I thought of a interesting idea that I’ll probebly try for my next project, I’m going to render an animation using opengl, and composit the particles and possible volumetrics in after the fact, this is mostly just a challange to myself to see the kind of results I can get with opengl rendering, now to the question part.
Does anyone here have experience with this approach and any possible tips and tricks? I’m no slouch when it comes to texturing and lighting or animation for that matter I’m primarily curious about the compositing aspects of this approach.
Ill be doing a lot of testing, in the meantime any advice would be helpful.
I’ve just finished doing this during my show reel development, to check the animation timing against the background and to see how the BG renders in terms of lighting and such. I rendered @ half scale (full size is 720p) so it wouldn’t take forever. The BG is mostly textured planes, so that rendered in full fairly fast, but because of the lighting setup, including the animated character really pushes the render time up. So I did OpenGL of the figure animation after cranking out a full-rendered background plate.
If in the Render context you set the Shading>Alpha: option to Premultipled, and render to RGBA PNG format, you can put the resulting OpenGL animation image sequence in the VSE and slop-comp it over the rendered BG using the Alpha Over Effect strip.
In general I didn’t use the Textured Solid mode because my textures look very odd without the full Material shading, but you can do this if your material/Texture is fairly simple and uses only a single texture. However at one point the character interacts with one of the tetxured planes, and in that instance, I rendered a black “card” with the chracter on Open GL, then the art for the card as a separate plate using Textured Solid, then added that to the VSE compo using an Add effect strip.
Upshot is that the OpenGL can be composited just like any other rendered element, depending on how you set it up. It’s a fast and accurate way to check on many aspects of a final rendered scene before actually spending time rendering. My graphics card isn’t GLSL-capable, so I haven’t tested that aspect. but for the most part, OpenGL is WYSIWYG, so you should be able to use it for some lighting and texturing tricks if your card can manage.
Thanks for the reply, I thought that glsl shaded opengl rendering were going to be aliased but apparently its not the case when you force AA with your card, the only thing I haven’t really tested is every render pass to see if it works, so far the ones that I have tried result in a black screen after composit so I’m assuming that it doesn’t. There are easy enough ways around this if the depths of objects are obvious, and I found out about the alpha last night (did notice a bit of white around the edges) overall I think this should be do-able and if the textures are nice enough the results really shouldn’t be that bad at all after compositing the stills.
I know this would normally be used as a previz animation but I thought it would be fun to make an animation focused around glsl. I will have to composit certain effects after the fact (particles and volumetrics) I’m mainly doing this as a fun excercise, if the results are good enough maybe I’ll do more animations like this every so often.
Using the Premultiplied option in the Render>Shading section should eliminate the white wire edge. I don’t think you can use OpenGL with Render Layers, though, didn’t think you were trying for single-pass compositing. All the passes you can split out with Layers and Render Layers are rendering-engine generated, so they probably won’t be accessible with OpenGL, which bypasses the normal rendering pipeline.
Strange I could have sworn I turned premultiply on last night and still could see the white edges, I just assumed it was result of forcing AA, apparently I should wait until I have plenty of rest before I perform tests.
I pretty much figured that the render layers wouldn’t work with opengl but it wouldn’t be very hard to do a final composit with alpha png, I suppose I would have to be willing to sacrifice certain render passes, if it becomes too much of a problem I’ll just render more with the regular pipeline and do as much as possible with glsl rather then primarily glsl.