|French Translation by Pilou|
This is the first in a series of articles which will go through how to produce realistic renders. Each article will focus on a particular aspect or rendering, and how to use Podium to deal with each of these aspects in terms of creation of a realistic image from a very simple scene.
If you are still reading, you will not only be interested in creating more realistic images of your SketchUp models, but you will also have a reasonable attention span! Some issues benefit from a little detailed explanation, but in all cases we’ll try not to go into any more detail than absolutely necessary.
It is aimed at the beginner to intermediate SketchUp user. It assumes you will have a degree of familiarity with SketchUp, and will be using groups, components, and drawing to scale. If you aren’t using these features in SketchUp, you need to learn how to use them, they are essential to effective modelling and rendering. It will take you from creation of a simple model, to rendering a fairly realistic image, by covering all the fundamental aspects of rendering.
First of all, what do we mean by ‘photorealistic’ rendering? This question is not as easy to define as you would think. At its simplest, one could define it as the creation of computer-generated images which look like photographs. This is very, very difficult indeed. It depends on accurately simulating the appearance of material properties, geometry, lighting, and camera effects. There are some excellent examples here.
The basics to creating great renders are to use high quality textures, an appropriate level of detail, and to set the lighting up correctly.
If we first consider materials, you might think that objects only have a very few basic properties like surface roughness or smoothness, colour, transparency and reflectivity. This is only partly correct as we will see later.
With geometry, real objects have lots of irregularities, some are immediately evident, like warping, cracking, bending and misalignment. Others are very subtle and might not be immediately apparent if you aren’t looking specifically for them. The most obvious example of this is edges. Most people will model these as simple extruded shapes with sharp corners. If you look at a table for example, where the top meets the edges, there will almost always be a slight rounding. It’s not sharp like a knife edge, there is a small curve. If you look at Figure 1 below, you’ll see that both the table and the tissue box don’t have sharp edges, and there is a soft blurred or highlighted edge instead of a sharp one.
Figure 1. example of soft edges
This changes the way the light falls on and reflects off surfaces. Little details like this can make a noticeable difference. Simulating this accurately would be very complex and time-consuming, but there are some tricks to getting great results without overdoing the detail.
With lighting, there are two basic types, direct and indirect. Direct lighting comes straight from the light source, indirect lighting is when the light bounces off surfaces onto neighbouring surfaces, causing shadows and illumination where you might not quite expect to find them. Figure 2 below shows both direct and indirect lighting.
Figure 2. Direct light
Figure 3 below illustrates indirect lighting very well. There is no direct sunlight at all, but the indirect light from the sky is bouncing down the deep narrow gap between the buildings and illuminating the walls, the street and even the underside of the balconies!
Figure 3. Indirect light
If you want to create really good renders, you need to balance direct and indirect light.
Finally, with photographs, there are particular features that are introduced by the camera that influences the final image. For example there might be lens blur, lens length or camera flash.
Simulating these things accurately involves a fairly detailed knowledge of each element, which is a lot of learning, yet people produce supposedly photorealistic images without all this detailed knowledge. How is this so?
To go back to the question posed earlier, “What do we mean by photorealistic rendering?”, the short answer is that in computer generated imaging, the term ‘photorealistic’ is relative, and is generally used to refer to the creation of images that look realistic as opposed to identifiably computer-generated. The main SUPodium home page movie rather conveniently shows this very clearly and Podium has been designed to make the creation of these types of image as quick and easy as possible.
The next article will deal with the creation of a simple scene for rendering, which will be used to explore the principles outlined above.
|Back to the Top|
Figure 3. Render of textured model
|Back to the Top|
|Back to the Top|
This, the fourth installment in our series of articles on creating realistic renders deals with modelling. We have already covered the basics, lighting and texturing, this time we are covering the the level of detail in your model, based on a very simple scene.
To reiterate the points made in previous articles, this tutorial is definitely not about creating a render which is indistinguishable from a photograph. That is difficult and requires a much more in-depth approach, and much more attention to detail. These tutorials deal with creating convincing images, to start people thinking about how to develop their skills to improve their Podium renders.
It ought to be obvious that you can’t start out with the very simplest of models, texture them nicely, set up the lighting properly and you will end up with something that genuinely looks real.
The Podium Browser makes it easy to set up your materials, and add good render-ready components, however this doesn’t mean that you don’t need to put in much effort with the base model!
It’s not particularly perceptive to say that there is a balance to be struck between adding massive amounts of detail to everything, and adding hardly any. The closer you are to the model, the more detail you need. Similarly, objects further away from the camera need less detail. 3d games engines have components which have different levels of detail, which are switched on and off based on the distance from the camera. The fully-detailed trees with all the leaves and branches turn into 2d billboards when they are distant.
When working out how much detail to add, you also need to think about exactly what you want to achieve. The more realistic you want to get, the more detail you need to add, but the longer your scene will take to render.
I think that for most people, a simple principle is to use only as much detail as you need.
It takes a little time to work out what this is, and it is different for most people. Generally people tend to render a particular type of scene for a particular purpose, so once they get an approach they are happy with, they will tend to use the same approach for every scene.
This article focusses on lots of little things that people miss.
This is the scene we are trying to copy.
Figure 1. Original source photo
This is the scene we ended up with at the end of part 3.
Figure 2. The pool scene after texturing, lighting, and adding the background in post-processing
It’s a reasonably realistic starting point, so let’s add some detail and see how we get on.
First of all I’m going to round the edges of the pool surround. In reality, not many surfaces have really sharp edges. If you think about it, sharp edges generally occur when materials are cut and/or ground. Wooden items can have sharp edges when cut, but you almost always have a slight rounding applied to them to prevent the sharp edges splitting. Items that are moulded, almost always have rounded edges, because most moulding processes aren’t accurate enough to get really sharp edges. Cast glass tends to have smooth ground edges to prevent people from cutting themselves. You’ll find the same thing on plastered walls.
The significance of this for rendering is that slightly curved corners reflect the light in a subtly different way. You don’t necessarily notice this unless you look for it, but the softness and realism of the edge highlights will make a difference in virtually all of your scenes.
The image below shows a new component for the browser, the Ikea Hol table.
Figure 3. The Ikea Hol table component, modelled and rendered simply
This component has been built as simple extruded surfaces with holes, and the texture applied to the whole component. It looks okay, and if you see how it looks from a distance, you can’t see that it’s not very sophisticated.
Figure 4. The same component rendered from a distance
Render time with this simple component is pretty good, however the close-up view isn’t good enough for a realistic render.
The first thing we need to do, is work on the texture. In reality, the wood grain virtually always runs along the long direction of a component. You can see the real photograph here. You can clearly see that the table is made up of individual pieces of wood, fitted together. Look at the direction of the wood grain. Let’s modify the component to show this.
Figure 5. The same component modelled with individually-textured components
This is a big improvement, and render time has increased, but we can do better. Look at those sharp edges at the top and on the corner. Let’s soften those. I use the Roundedge plugin by Fredo downloaded from the ‘Plugins’ section of the SketchUp forum. This makes adding bevels and curved edges very quick and easy. It works fine with Podium and it’s one of my most frequently-used plugins.
The trick is to make sure you don’t add too many polygons unnecessarily. I have applied a 0.5mm edge with a single segment, (effectively a chamfer) along the vertical corner member.
to the lid I have applied a 2mm rounding with 2 segments. The geometry has been smoothed to make the curves look more natural and help with texturing.
I have also added tenon joints at the top, where the timber sections join. The render for this is shown below.
Figure 6. The same component with rounded edges
Render time has increased again, but the component looks much nicer. each of the elements is a component, which keeps the file size down for the Browser. We could improve the component even more by creating more variations on the slat members, and changing the texture, so that the wood grain is more realistic and less regular. You can particularly see the subtle highlight on the top edge of the lid and the tenon joints. With a little more texture variation, the component would look even more realistic.
Having demonstrated the point about texture variation, detail and curved edges, let’s return to our test scene. Concrete generally has slightly rounded edges, because it tends to crumble when too thin, and ‘normal’ construction concrete doesn’t have sufficient strength to keep really sharp edges.
I have softened the edge of the pool and the copings around the roof. Let’s also add some skirting boards internally, and some door handles.
The next stage is to add the furniture. I spent a little time building some similar items, and found a cactus model on 3DW which I textured and modified. The petals are dynamic components I built and distributed manually in a similar way to the photo.
Figure 7. The base model with more detail and additional furniture components
This scene has been rendered with the QMC preset, with the Podium physical sky on, and the transparent png background option. It’s not bad, but it we look at the door threshold and the amount of light inside the room, they are not quite right.
To increase the light inside, there are 2 options, cutting openings (because we don’t know how where the light comes from in the actual building in the photos) and adding LEMs. I have chosen the first option. You will also notice the area of wall above the lower roof on the right is a little overexposed. To fix this, I have darkened the colour of the roof, which will bounce less light off it and help to reduce this. You can download the modified model from here.
Figure 8. The same scene with additional lighting introduced into the interior space
This is a definite improvement. However we still need to add the sky in post-processing. At the same time we’ll adjust the brightness and contrast, and resize the image, because the QMC presets introduce noise or ‘graininess’.
The final image is shown below.
Figure 9. Final render with post-processing
We have adjusted the colour balance, saturation and contrast, and added a border.
The image could undoubtedly be improved even more with more sophisticated post-processing. With a solid understanding of blend modes, adjustment layers you can significantly improve any image. I’m not very good at post-processing, so I tend to limit this to the basics!
It’s not an exact replica of the original scene by any means. In particular, the wall texture and floor textures are much cleaner, but it’s a decent image, and shows how that to make your images that little bit more realistic, the better they get, the harder you have to work to improve!
In the final article, we’ll try to create a convincing night version of the render.
|Back to the Top|