Making Of 'CG Painting'
Hello, my name is Vadim Valiullin, I live in Russia, Ufa, and I am currently a student of the 5th course of Bashkir Pedagogical University, studying design. I want to share with you an article about the creation of the work, "CG Painting". I would like to note that this article reflects my point of view and is not an axiom. 3ds Max 8 (2008), Mental Ray and Photoshop CS3 were used in the creation of this 3D image.
I had an idea for this work one day: I thought that I had not been drawing using traditional methods for a long time, and I was thinking about how fine arts have practically passed into digital format. I remembered how I painted from life with pleasure a few years ago, and I imagined a situation in which I could take the computer out into plain air. The idea seemed pretty crazy, but funny. I thought about it constantly and eventually I imagined the scene visually; after 2 months I decided to turn the idea into 3D. So I had a distinctive idea of what do I wanted to achieve at the end; there was no need to jot down the concept or a sketch, I used only my imagination. It took about 1.5 months to complete the work, all done in my free time (Fig.01)
This work is not complicated in terms of modelling; all the models are rather simple and were made from primitives and splines with Editable Poly. Using the rich collection of Editable Poly tools I have not met any serious problems in modelling. At the same time, with modelling, I did the texturing and shading, but I will tell you about that later. In my opinion, we have an opportunity to control and correct details and not to waste time, to get rid of problems and to use recourses more efficiently. At the beginning, we should keep to the shapes and the peculiarities of the surfaces of the objects. If it is a wooden board we should curve the geometry, because ideal and equal boards do not exist. Also, the chamfers will have different thicknesses and irregularities. Such small, and at first sight imperceptible distortions of geometry will give a more natural feel to the work. Here are some screenshots of the geometry of the wood (Fig.02). There is nothing difficult in the modelling of these objects so I will not pay too much attention to it here.
If it is necessary to model metal then do not forget that metal usually curves, it does not break and burst like wood and so it would be good to make some deviations away from symmetry and ideal forms when poly modelling. It is also necessary to remove chamfers on metal objects or it will be difficult to achieve normal highlights, despite even the most advanced shaders. If we have to model some all- metal objects, the more differences we have, the better (but it is not good for advertising visualisation). Also, the cast objects take more numbers of faces or they will look rather rough, like forged products, so the Mesh Smooth tool is therefore necessary, especially when they are visible from a close distance (Fig.03).
As for the geometry of plastic, it will be better to stay equal and only to add chamfers (plastic is more round on the corners than metal) because plastic will burst and will not curve and distort. Moreover, plastic is rather flat and it is practically impossible to notice changes; the abrasions we can imitate with shaders. The wire in some places is not ideal, some objects are not whole and somewhere it would be possible to save faces. My aim was to do everything fast and not to use too many faces. In Fig.04 we can see that sharp and direct lines reflect the shape of plastic products.
I modelled the paint box looking at photos; I had to think up some of the elements because of the absence of certain references. Nevertheless, I tried to make it all true. I then supplied the paint box with the monitor, and drew a preliminary sketch as a texture on the screen (Fig.05).
It was necessary to fill the paint box with small things, so I will tell you about the ideological components and design of some of the nonexistent objects. Certainly, I thought about their design myself, looking at real-life objects. It was simple to do it with the paint-box, the monitor, the system block, etc.; I based them on existing products but replaced some brands so as not to advertise anybody (I'll tell you about rebranding later).
I decided to make a palette; I didn't want it to be your usual palette, but CG. I therefore decided to mix a graphics tablet with a palette used for traditional painting. It is always easier to model when the sample is near and we can touch it and look at it during doubtful moments. Basically, there are two kinds of palettes: rectangular and oval. It is not worth mixing rectangular forms of a tablet with an oval palette; everything should be harmonious and convincing. I therefore took the rectangular palette as a base; I did the same with the brush (Fig.06).
Then it was the turn of the palette knife, which is used for mixing colour in traditional painting and for applying oils to canvas. I decided to take the form of the Photoshop tools icons as a basis for the form of my palette knives, such as: Sharpen tool, Blur tool and Smudge tool (Fig.07). At the same time I modelled and textured a set of small things for the subsequent filling of empty spaces. All textures of wood on the handles of the palette knives have different shaders and mapping; the brushes are the same. This promotes the best perception of the work. My rule is not to use full copies of objects in one stage if they will be seen from one shot and if it is not necessary for the task.
The next stage was the paints - gouache and oil. I decided to make them on the basis of CMYK (for print) and RGB (monitor display). Making textures for the paints I replaced all colours and added corresponding inscriptions (Fig.08). Moving forward, I would like to say that all chamfers on the boxes with paints were isolated by texture. In some places I drew them in Photoshop, as it helped make the work look more convincing.
I designed water colour paints, too, taking the PANTONE colour system as a basis for these. Unfortunately, I couldn't think up anything logical for the solvent and did not want to go into a complete fantasy, so I went for the most simple and obvious solution and added "For digital artworks" - not specifying anything in particular (Fig.09).
Using my modelled brushes, paints, small things, etc., I began to manually fill the paint box with them. It was possible to use a reactor or a scatter for this purpose, but then there would not have been the full control over the position of each object that I wanted, and as a result everything would have had to be corrected. I spent a lot of time filling my paint box, but at least I was able to control the position of each object. In order to the objects look more natural, I tried to do it so that the copies of objects did not lie under a corner or on the same side as the viewers. I paid special attention to this stage of the artwork, knowing that the final result would depend on how harmoniously I scattered all objects (Fig.10).
When paint box had been filled, I started the configurations of larger objects. I also changed trademarks and names on visible objects not to be engaged in advertising. All trademarks and names are fictional (Fig.11).
For the preparation of textures, I basically used my own photos and drew them in Photoshop. In some cases I partially used photos from the Internet and my library of textures. Most of the objects were going to be smeared with paint, and so to draw textures with paint on each object separately was not rational and would have been very time-consuming. I therefore created in Photoshop the image of spots of paint in the high-res with an alpha-channel. I could then impose this image, change its transparency, contrast and saturation on any texture, whether it was a texture for wood, metal or plastic. I needed to make sure that it was not so strongly pronounced as to cause tiling on different types of textures (Fig.12). (I cannot tell you how I created the texture for the wood or plastic because it was a serious theme and took a lot of time.)
For many objects (where there are some inscriptions) I made unique textures. I did not unwrap and did not use Unwrap UVW, as at this stage of the work there was not such a complex object. For texturing there would not be enough standard forms of UVW mapping; in some cases I used different mapping and different structures for the sides of the objects and it was quicker to do than high-grade unwraps. Thanks to this alone, I saved a lot of time, and you cannot tell in the final result that I took such shortcuts.
To further explain, I simply adjusted the size of textures to the size of objects. On practically all identical or similar objects the textures had different shaders and different mapping. Drawing textures, it was necessary to consider the fact that in corners and under ledges, objects, as a rule, would look shabbier and darker from the accumulation of dust and dirt. I gave more light to the scene and also blacked out some areas in Photoshop, using various brushes; in the most cases the standard brushes were enough. These nuances are visible on the textures of paints, the monitor and boards, etc. Textures should be more or less neutral in terms of lighting, and there should not be any strongly pronounced shadows on them, otherwise, during rendering, the different light exposure will give an artificial look. On the texture of the paints a shadow is visible, which I transformed to rust because the unclear shadow can confuse viewers and look more like real rust (Fig.13).
This is a rather difficult theme and there are many differences, but I will try to briefly describe my approach to it. At the stage of shading, it was necessary to consider real characteristics of the different materials. Unfortunately, when adjusting materials to enter the valid physical parameters, the result is not always satisfactory. There is nothing surprising about this though, as a lot depends on the adjustments of light, the distance from light sources to objects, the corner of an inclination of the camera, etc. So to adjust materials we have to use an experimental method. It is desirable to know what functions parameters have. I basically use standard materials in most cases, when they are enough. In my opinion there is no need to use advanced shaders in all cases just because they are "advanced". They should be only used to achieve an effect when using standard materials is impossible. Such an approach helps to save a lot of time, both during the adjustment of materials and during rendering.
As already mentioned, I use Mental Ray as my renderer, and in most cases standard shaders. This was convenient because I could adjust materials rather quickly, using Scanline for test renders. As a rule, when rendering using Mental Ray, materials look approximately the same and even better. It is a more convenient way to test them than to adjust shaders and make advanced renders which make you wait a long time for the results of the tests.
When creating shaders of wood, it is necessary to consider that wood has practically no highlights, except for varnished wood. Metal has strong highlights that can reflect objects around it; if there is nothing to reflect then the metal will look poor. We have to create the surroundings, or if possible apply textures on the reflection or use HDRI. There are two types of plastic: matte and glossy. It is better to do matte plastic, paper, rubber, using SSS for glossy plastic. As a rule SSS is not required, although there are lots of different variants. It is also important not to forget about bump, displacement, specular, etc. If we use them right then it is possible to achieve very good results! It is desirable that the bitmap on them differs from diffuse, but for average and distant planes it is enough to use diffuse on bump, specular, etc. Materials will behave differently under different lighting conditions. After the final lighting stage I had to correct all shaders in which SSS and reflection were applied. In Fig.14 are screenshots of some materials from the material editor.
Adjusting the lighting is one of my favourite stages. Great lighting can improve an adequate image; bad lighting can ruin a masterpiece. To save the time I decided to adjust lighting using grey materials, since for the adjustment of good lighting it is necessary to have a number of preliminary renders.
It was necessary to imitate sunlight and I decided to use the Daylight System, but I had to replace 3ds Max 8 with 3ds Max 2008, as 3ds max 8 was not enough for a good result. When adjusting sunlight, it was necessary to remember that the light of the sun has a warm shade, but all shadows have a cold, bluish shade to then, simply because shadowed areas reflect the colour of the sky. It was also necessary to consider the fact that shadows from the sun become dimmer the further they are from an object projecting a shadow. It is impossible to use absolutely precise shadows; this is a very rough similarity of what it would be like in reality (Fig.15). In 3ds Max 2008 the Daylight System is realised well enough and the adjustment of lighting has no special problems.
When the foreshortening had been chosen and the light was adjusted, it was possible to start filling a stage with small things and study the details of the surroundings. By means of a graphics tablet I made some approximate sketches and then modelled on top of them. I considered such an approach effective enough - it also saved time. During the final stage it was not necessary to think and guess about how to fill empty spaces; I could rather quickly draw lots of sketches and choose something from them to build upon. Besides, it was also possible to solve all compositing and colour problems in parallel (Fig.16).
Afterwards, I began to do the grass and the background; for this purpose I transformed all objects into a grey material because so many resources were demanded for this part. I decided to make the grass by means of the Super Grass plug-in. I made plenty of preparations for the grass, with different heights, thicknesses and distribution over a square metre. I then converted all of this in Edit Mesh, optimised by means of the Optimize modifier and manually spread it out how, in my opinion, low grass on hills would grow under the influence of constant blowing winds. This was probably not the best way of creating grass, but it allowed me to supervise many aspects well enough (Fig.17).
The grass and background ate a lion's share of resources, so I had to render the final image in layers: the paint box, grass and background separately. It took about 2 days to render it in high resolution, with all layers for all foreshortenings. I rendered it with Depth of Field, and also with Global Illumination and Final Gather, since together they give very good results.
Finally, the turn of post-production came, for which I used Photoshop. I did the post work without rushing, as by having a break from the image and getting a fresh perspective on it, it was possible to see some areas that were lacking. Before the final stage of post-processing I looked at a lot of high quality three-dimensional and two-dimensional work, and also a couple of films from Pixar to distract myself from my own work. In my opinion, this method allows you to raise your taste in art taste up a level, and as a result helps you to notice differences in your work which you will have missed when working on an image for a few weeks. When you're working on something for a long time and without breaks, it becomes more difficult to notice problems.
To complete this article, here are some images showing the stages of post-production, along with the final image (Fig.18 - 19). Thanks for your attention.