3D Process / Post Process: tricks revealed!

Many of you have asked for my post production process…so here it is.  I typically use AE (Adobe After Effects) for my post work, but for this post I’m demonstrating in PS (Photoshop) because most people use PS over AE for stills.  However the principles apply to all software.  Also, I’m trying to keep it “out-of-the-box”, rather than show a lot of plug-ins.  I always suggest learning the techniques with the software, then after understanding how to create them, go get the plug-ins to make your job faster-not better.

Some of you already know, but I do most of my modeling in SU (SketchUp).  Not because it’s better than Max, but I find it super fast for building design process.  Because we do architecture, SU is very good with boxes and simple shapes.  If you want to get into character modeling, Max or Mudbox is better for that.

I also start materials and texturing in SU.  I find SU super easy and fast for texture layout.  If I texture most things correctly in SU, I can almost avoid the UVW layout process in Max entirely.  The materials in SU are nothing special, just place holders really for the maps I want to use in 3ds Max.

Using Max 2011 has been great, because I can import SU files straight from Max without needing to export models from SU.  This new importer in max is incredibly powerful.  It respects instanced components from SU, remembers UV texture position from SU, and converts SU materials to A&D materials automatically….amazing tool.

During import I opt not to import the SU cameras.  I prefer navigating in 3ds Max to get my actual camera angles.  This is where I add any entourage from my library of 3ds Max models.  Furniture, cars, plants, trees, etc. all get added here.  Then I begin texturing.  I swap out SU textures for better texture maps.  Or sometimes I replace a texture map with max procedural maps like tiles and gradients.

Some of the general settings for my A&D materials:  If I can keep glossy samples to 8 then I do.  Under Special Effects, I usually turn on Ambient Occlusion, and set the distance to 3’.  Under Advanced Rendering Options, I make sure that backface culling is un-checked so I can render both sides of a mesh.  If I have a single plane of glass then I check Thin walls.  If my glass is a box or has thickness, then I leave it as solid/thick.  I find the real magic behind getting realistic renders lay in the material reflections.  I usually have a reflection map that drives how much reflection takes place.  In this tile material, the grout lines are black meaning no reflection, and the tile is more white meaning a lot of reflection.  I use the same map for a bump effect.  I almost always have my glossy reflections lower than 1.0.  For this example I have it set to 0.4 – meaning the reflection is scattered at 60%, in this case with 8 samples.

Next I go to lighting.  In this scene I have 1 Daylight System, 309 photometric lights, and 5 mr sky portals, for a total of 315 lights in my scene.  This many lights in a scene would typically be brutal.  However, for my photometric lights, I opted to use Point for my Shadow type.  It doesn’t look as good as the other options (Line, Rectangle, Disc, Sphere), but renders much faster than the others at their default setting.  For every shadow that Point renders, the other options render 32 samples per shadow.  So this is a big render saver.  For my photometric lights, I usually use the default light levels, and switch to photometric web using an IES file for the distribution.  As for the mr sky portals, I try to limit their use to just where the large windows are.  Render times take a big hit from mr sky portal shadows as well.

This image is what the render straight out of max looks like, known sometimes as the beauty pass.  I render inside models with GI & FG.  Surprising to me, I rendered this scene with the default settings for both GI & FG.  I was reasonably happy with the results.  I did get noise near some of the clerestory windows, but I was willing to live with it.  I left all of my lights on, then calculated GI, saved it to a file, then rendered FG from each camera adding onto the previous FG map.  Before rendering the final renders, I had 1 GI map and 1 FG map for the entire scene.  This made it easier for me to switch cameras and not have to worry about changing light maps.  The GI map ended up being 154 MB. The FG map was rendered at 50% from the final renders at 800 x 400 pixels, and ended up being 34 MB for all 11 camera angles.

This is the Ambient Occlusion Pass.  If you want to know how to do this, check this post out.

This is a flare pass for the lights.  This can be created in 3ds max, or in post.  I usually create this image in PS, it’s faster and gives me greater control on what the flares look like.

This image is a dummy people pass.  I rendered this one out to give me correct scale for adding people in PS later.  This way my people won’t look like giants or elves when I scale them.

These are the people that replace the dummy people.  To each person I add motion blur, reflections, shadows etc.  Then I save this as a .png file and add it to my beauty pass.

This is a volume pass that I render in 3ds max.  It is created using the Parti Volume Shader.  I then add it in PS and tweak it to the right look.

Lastly, I render a Z depth pass.  Depending on the rendering I sometimes use this.  If there really isn’t an object in the foreground then often I don’t use this at all, and rather just manually blur the edges of my image.

With all of these passes combined in PS, AE or other compositing photo/video editing software, you can take your original image and turn it into something much stronger visually.  This quick video should give you an idea of how I add all of these elements together using color correcting, layers, levels, to transform a raw rendering into a finished rendering.

If you would like to see all of my renderings from the KHS project, click here.