Part II - Lighting & Rendering
This article will cover the technical aspects in the making of my piece Starry Night. I hope you will find this project overview useful. Maybe you'll even learn something new and be inspired in your own work. Part 2 of 3 parts covers lighting and rendering.
Lighting is an important aspect of an image that's crucial to the mood in it. Many people use the well known three point lighting technique when setting up their lights. I use that technique from time to time too, but sometimes, as in this image, I don't stick to it 100%.
I lit this scene without any back light (instead I prefer using normal-color ramps on the appropriate materials), and the fill light very close to the key light. So ultimately it can't really be classified as three point lighting at all. It's just what I felt I needed for the scene.

I nearly always use area lights (or spotlights with soft buffer shadows). Because if there is anything that looks bad in my opinion, then it's hard shadows (they should only exist in direct sunlight - and can/should in certain situations be blurred there too). Using area lights gives me soft shadows, but longer render times on the other hand, so that's also worth keeping in mind.
To enhance the fake subsurface scattering I wanted to have the ears glow a little bit as they would when being lit from behind. Not much, just a little bit. I added two small spotlights, one for each ear, to illuminate the ears slightly.

To achieve even softer shadows I used Ambient Occlusion along with the lights. It fakes Global Illumination (provided in it's true form by higher-grade renderers with long render times) and gives the whole lighting a much softer appearance. Many people use Ambient Occlusion too rarely. I've seen good modeling, textures, materials along with a fair lighting setup fall apart, simply because it needs that last touch that Ambient Occlusion delivers. It's really a shame. One thing to keep in mind is that Ambient Occlusion will greatly lengthen render times.

I was going to do post-processed depth of field (Lens focus blur in Photoshop) for this image, so I quickly realized I had to render the scene out in render layers, and composite it all in Photoshop afterwards to let me control the amount of blur for each layer.

Rendering in layers is often preferable. It enables you to do a much wider range of manipulation in the post-processing face, letting it truly flourish and come to it's right. But more about that in Part 3. I split the scene into 4 layers: Background, Body, Head and Hair. The reason i split the hair from the head is that it needs some special care in the post-processing phase to look right.

Render passes are images that contain different isolated image data from a scene, enabling you to have even greater control in the post-processing phase. Normally you output many different passes like diffuse/color, specular, ray mirror/reflection, Ambient Occlusion, Light and so on. Render times can often be shortened by splitting the render up in passes.
For this particular image I only needed a few different kinds of passes: a couple of z-depth passes and a specular/ray mirror pass (on the eyes only) to make the star glow effect that gives a nice touch to the eyes and emphasizes the reflected stars.
The reason I still haven't done much work with passes myself is that Blender as of this date doesn't have an automated render pass system. Doing passes in the current Blender release involves a lot of copy'n'paste work and tedious material work. This is not very efficient at all. (Note: A nice render pass system has since been implemented in Blender).
I copied the entire head render layer and adjusted all materials but those for the eyes to be completely black. The eyes themselves where adjusted to black too, yet they had ray mirror reflections and specular as the only things showing. When rendered it looks like below.

A z-depth pass is an image used to control how much focus blur should be applied to an image. In real camera objectives, focus blur gets stronger the further objects are away from the camera. Each pixel in the z-depth pass has a value between black and white. The lighter the pixel is, the further away from the camera it is, and the darker the pixel is, the closer it is to the camera. This image data is then used to calculate how much focus blur is needed, thereby replicating the focus of real life cameras.

In Blender 2.42, the new compositing node system allows you to do z-depth passes much much faster and more integrated than ever before. The node setup below can be expanded to obtain the focus blur effect within Blender (using the Blur node), but also - as in this project - saved as an image file to be used later in an external image application.

The render layers and passes are saved to 2400x900 pixel png's (with alpha channel data) to be composited later in Photoshop. It's good to be sure to have a decent native resolution - you never know what an image might be used for. Print, display on HD screens, maybe even billboards if you're lucky! (Though if that's the case you should go with an even higher native resolution than I did). At the end of the day a fair amount of different image-types had to be output from Blender on this project (7 to be exact).
This ends part 2 of this article. Please proceed to part 3 where I write about compositing the render layers and passes together, image enhancing/adjusting and the final little touches that add flavor to the image. Also if you haven't read part 1 of this project overview yet you should consider doing so.
Making of Starry Night | Part III: Compositing, Adjustments & Final Touches
Making of Starry Night | Part I: Modeling, Textures & Materials
//Mathias
Have you found something not making sense, spelling mistakes or typos? Or have you got any questions that you feel I didn't answer? Contact me, and I'll try to see if I can help you/improve upon this.
<< Back to Resources