Making Of 'BMW C1: Urban Trespasser'

Introduction

Wikipedia says:
"The BMW C1 is an enclosed scooter or feet forwards motorcycle, which was manufactured by Bertone for BMW. It was intended to combine the advantages of two-wheeled motorized transportation with the strengths of the automobile. The idea was to offer the convenience of a scooter or motorbike but without many of the associated dangers or hassles. BMW tried to add passive safety and car-like crash testing to the scooter. It claimed that in a head-on collision, the C1 offered a standard of accident protection comparable to a European compact car. Since going out of production, C1s have found new homes with collectors and as pit-bikes for some of the Formula One Grand Prix teams."

In my non-commercial projects I try to select a subject and components that will make use of my skills to a maximum, and I set myself tasks to solve using new methods. In particular, I was interested in industrial and automobile design, as I have no experience in these fields, and I felt that work of this nature would grant me new opportunities to develop my skills. Considering this, C1 proved to be interesting! I bumped into it accidentally, looking through Google Images. The style of this concept seemed fresh and unconventional, it stood out of about 20 other images on the page, and in the next moment I was looking for its blueprints.

Modelling

In the beginning I had no idea how this project would look in the end. One thing was clear at that moment, though: the final image had to satisfy the requirements of A4-A3 printing and render options were to be for high-res. The motor scooter itself was meant to be the main subject, and the most worked-out part of the overall composition, and so further planning was done based on the most advantageous views of it.

At this first stage, my main goal was to make an authentic model, and it was at that moment that the first problems arose; the one set of blueprints I found was poor in detail and incorrect - the placing of some of objects didn't correspond with one another in the different views, and so I could only use them to form the approximate cargo body proportions only (Fig.01).

The modelling sequence can be defined as assembling the car body first and foremost, constantly fitting its parts to one by one, forming the silhouette and the main proportions and then adding the details to basic forms corresponding to the references (Fig.02).

My technique itself was nothing special - simple poly modelling done in the majority of existing 3D modelling software. As usual, I began by building the form from one polygon; I created a plane with one segment, converted it into Editable Poly, switched to the edges editing level and moved them by pressing the Shift key and "pulling out" remaining polygons. If I needed to join model parts, I switched to vertex editing level and connected them using Weld, Target Weld or the Collapse command. If I needed to cut out part of the model or add more edges in places where the grid was not dense enough, I used the Cut and Slice commands at polygon editing level, or the Connect command at the edges editing level.

If the topology of a model satisfies me, but its form and proportions do not, I use Free Form Deformers. I also have a habit of using Edit Poly as layers, like in Photoshop or ZBrush, to have the opportunity to compare the things as they are and were, or to undo it at any moment (Fig.02 - 06).

Fig. 02

Fig. 02

Fig. 03

Fig. 03

Fig. 04

Fig. 04

Fig. 05

Fig. 05

With this project, Edit Poly's standard features were surely enough, but it's always useful to remember that they can be extended with scripts like Advanced Poly (http://www.ap.nine.ru/) or PolyBoost (http://www.polyboost.com/).

After the assembly was done, I took 4 views that seemed interesting and descriptive to me (Fig.07a - d). On the base of these the whole scenery was created, as planned (the last 4 views which made it to the final set of images were selected by taking into account the surroundings after the modelling stage was finished).

Fig 07

Fig 07

Fig. 07a

Fig 08

Fig 08

Fig. 07b

Fig 09

Fig 09

Fig. 07c

Fig 10

Fig 10

Fig. 07d
At this stage I intended to use studio renders only, but I wasn't sure if this variant suited me ... I had to complete some test renders to get a rough idea of the result. To achieve this, I set the test lighting with minimal settings. I selected HDRI to be well colour-balanced, because I wanted the colours on the render to remain the same as they were set in the Material Editor, without adjusting White Balance in the Camera settings - it was easier to make changes in the materials this way, and be sure they would look natural in different lighting situations (Fig.08a - d).

Fig 11

Fig 11

Fig. 08a

Fig 12

Fig 12

Fig. 08b

Fig 13

Fig 13

Fig. 08c

Fig 14

Fig 14

Fig. 08d
I then created some simple materials and rendered a couple of rough images of the studio (Fig.09a - b). I wanted to make two variants of lighting to show how the model would look during the day and night, and in this respect I was disappointed with the results. In the studio with a dark background the difference in lighting wasn't really noticeable, and the studio with a bright background looked plain and somehow unfinished. I therefore came to the decision to try the variant with surroundings - it helped me to create a complete and complex picture with elements from different areas of 3D design; I could check my skills at all phases of the process and get the maximum from the project. It didn't make sense to restrict myself by creating a background only for one view and to leave the others in the studio, as the whole 3D scene met the requirements of the model's presentation for each selected view, and could also be used in an animation for a show reel, if I wanted to do one in the future.

The subject for the background was defined at once; I've always wanted to create something linked with industrial culture, and here I was given the opportunity. At that moment I already decided to "dirty" the model a little. The industrial background in this case takes the role of an aggressive environment, explaining the stains on the hull by its existence, making some contrast with the C1's futuristic look and also allowing me to play up the black and yellow colour scheme of some dangerous mechanisms. All of this provided interaction with the model itself and its surroundings, making it organic and associated.

Fig. 09a

Fig. 09a

Fig. 09b

Fig. 09b

The next phase was to search for references for the environment (Fig.10). The work on gathering references was completed in the following stages:

1. I made a list of items I wanted to see in the picture and made some sketches of the scene with dummies made of primitives. At this stage I needed to set the volumes. I didn't want to make the surrounding structure too complicated and fractional, in order to keep the viewers' attention on the bike, so I tried to build the space mainly with large forms.

2. I looked for colours, materials and textures for the objects selected at stage 1.

Fig. 10

Fig. 10

Texturing: Materials

With the modelling work finished, all objects were scaled and placed, views were selected and the lighting was ready. So it was time for textures and materials.

All models in the scene could be divided into three groups by the complexity of their unwrapping, as follows:
1. Models for which one of the standard variants of UVW mapping was enough. They have simple form or they are some undistinguished details of a more complex object - mostly screws, screw-nuts and the like this, as well as objects in the depths of the scenery and poorly lit ones.
2. Models that need unwrapping on the base of standard mapping which demands correction "by hand", but without strict accuracy. For instance, when you need texture scales of different objects to roughly correspond to each other, or when you need to place some stains somewhere without taking into consideration the seams and texture stretching because the model's form allows you to hide it easily. Most objects of the scene are in this group (Fig.11a - e).

Fig. 11a

Fig. 11a

Fig. 11b

Fig. 11b

Fig. 11c

Fig. 11c

Fig. 11d

Fig. 11d

Fig. 11e

Fig. 11e

3. Models which needed unwrapping to be planned or made under specific requirements. These were parts of the hull:

3.1 Shroud. Lines of the texture needed to be parallel with seams on the hull; therefore the proportions and contour on the unwrapping were to be the same on the hull and its parts. These requirements, for instance, excluded the use of Pelt mapping (Fig.12a - b).

3.2 Tray. There must be some relief on the plate on the bottom of the hull. The easiest way to make this relief is by using Displace. As the relief's pattern should be balanced along the hull, I needed to use planar mapping on the displaced part to preserve its contour and proportions, and to build up the rest of the mapping around it, avoiding texture distortion and hiding the seam which was achieved by using Pelt Mapping, Relax and handmade vertex positioning (Fig.13).

Fig 23

Fig 23

Fig. 12a

Fig 24

Fig 24

Fig. 12b

Fig. 13

Fig. 13

The materials could also be divided in two groups, as follows:

1. Standard. In most cases these are enough to achieve simple tasks. The basic principle is to achieve the goal by playing with basic material settings, using  textures for the slots of Diffuse, Reflection, Glossy and so on, and corrections of these textures by means of the texture's Output menu (or by means of plug-ins like Colour Correct http://www.cuneytozdas.com/software/3dsmax/) (Fig.14a - b).

2. Composite. These are used in specific tasks, built up on combinations of different materials or textures by means of masks or Procedure maps. They have "branchy" hierarchy. A striking example in this work is the ground material.

Fig 26

Fig 26

Fig. 14a

Fig 27

Fig 27

Fig. 14b

Texturing: Ground

Formerly, all the textures were prepared in Photoshop. In the material editing I used finished images that were received from the Photoshop output. If I needed to edit the texture, I returned to Photoshop. Exceptions to this were in cases when only the contrast should be changed - instead I used curves in the texture's Output menu for it. It's a standard approach which lets you edit materials without complicating their structure, and in the beginning i wasn't intending to digress from using this.

To begin, I needed to define the resolution of the texture that I had to work with. There was no sense in making a unique drawing of the texture for the whole object, as it would be partially hidden by other models. Besides that, at this stage I was working in 32-bit WinXP and was restricted by 2GB RAM for a programme. I wrote "3gb" in boot.ini and it gave me a little space to act, but I still needed to use memory sparingly. So, when the unwrapping was done, I scaled it in such a way that 3ds Max's unwrap projection contained only these parts that could be seen in two views minimum. This way I got a square with the model of the bike in the centre of it - this square didn't cover all visible ground areas which led to texture tiling, but as it didn't fully appear in the field of view of any camera in the scene, tiling wouldn't be visible as the tiled part of the texture was outside of the screen, or hidden beneath the environment (Fig.15a).

Fig. 15a

Fig. 15a

With the texturing region determined I then moved onto the resolution. To define this, I used a 1000x1000 pixel texture for the object, with which I intended to imitate the ground and made some renders in final resolution, increasing the tiling until aliasing disappeared. The figures I got looked bad - the tiling measure increased to 12. This meant that I needed to use a 12000x12000 texture to get a decent picture of ground, but at the time 3ds Max crashed when I tried to use textures larger than 4096 pixels on the largest side.

One thing I could do was to assemble the needed texture from smaller ones inside 3DMax, like this:

- I took some textures of tarmac, rocks, earth, mud and sand

- I set the textures' scales to the scene scale by setting the tiling measure for each one of the textures

- I made lighter and darker variants for each texture

This way I got about 30 layers which I needed to blend with each other to "break" the tiling effect of each layer and to remove the synthetic uniformity it made, making the final texture more natural-looking and credible. Following the same principle I also made sure that the tiling effect of textures with the same resolution had different values, so that seams between the tiles on different layers didn't lie in the same places. Then, using the Mix map, I began to blend the layers.

In the beginning I intended to use procedure maps as masks, but I wasn't satisfied with the result and decided to use textures. I was limited in their maximum resolution as before, but as they were used to mark the sections of layer mixing, they could be stretched without risk of worsening the final image, as the scale of the layers remained the same. Cheating it is, and it led to some artefacts, but they were easy to remove with the Healing Brush in Photoshop (Fig.15b - d).

Fig. 15b

Fig. 15b

Fig 30

Fig 30

Fig. 15c

Fig 31

Fig 31

Fig. 15d
At this stage the scene was highly unstable; 3ds Max was ready to crash in response to any careless action, which is why the ground material was primitive - I placed the final texture in the Diffuse slot and finished with it. For displace I used a copy to have the opportunity to control the height and direction od the displace, without any changes in Diffuse (Fig.16).

Fig. 16

Fig. 16

' '

Lighting

I wanted the lighting in the scene to be mild, with degraded shadows - typical for cloudy summer's day. The light needed to be directed to delineate the model's form and to add some depth to it by contrast of highlighted and shadowed parts. For these goals, HDRI with a Dome Light were enough for me. Tests with HDRI met all the requirements mentioned, so I kept going with it.

I adjusted the Dome as a direct light (i.e. I didn't turn "store with irr.map" on). In this mode, Dome with HDRI gave more correct and detailed lighting than just an HDRI set as a texture for the Environment (Fig.17a).

I also tried turning on "store with irr.map" - all highlights in the scene were practically washed out, and to keep shadow details I had to twist out the Irradiance Map adjustments, which made render times longer and gave me even worse results.

As this scene was of an exterior type and had few places which were unseen and unreachable by primary lighting, the influence of secondary lighting was insignificant, so it was needless to adjust its settings whilst getting a realistic and pleasant look in a lesser time. Nevertheless, I turned "store with irr.map" on in the Dome's settings for night renders, but only because as a direct light there were planes that I used to imitate flashlights (Fig.17b).

Fig 33

Fig 33

Fig. 17a

Fig 34

Fig 34

Fig. 17b

Some more planes were placed inside the headlights. I turned off caustic in the global lighting settings to decrease render times, which is why the glass material with a Glossy effect let the light through badly, and the headlights gave almost no reflection to nearby surfaces. For night renders I decided to accentuate this effect, and for this purpose I placed more invisible planes at the outer side of headlights (Fig.18a - d).

Fig. 17b

Fig. 17b

Fig. 18a

Fig. 18a

Fig. 18c

Fig. 18c

Fig. 18d

Fig. 18d

Rendering

The render settings were based on LWF (Fig.19a - b). You can read a detailed description here:
http://www.gijsdezwart.nl/tutorials.php and here: http://www.chaosgroup.com/forums/vbulletin/showthread.php?t=33917.

Fig. 19a

Fig. 19a

 

Fig. 19b

Fig. 19b

Post-Production

Working with LWF often results in a final picture with low contrast and weakly saturated colours, which is why the main goal of post-production is to strengthen these aspects. I have no general algorithm of actions; for each image I have a specific sequence. Below is just one of the variants, but if I work with LWF I always begin in the same way: I save the results of the render in 32-bit format (.hdr or .exr) to work with it like with an HDRI (to apply Tonemapping in Photomatix or in Photoshop, for instance). All of the whys are covered in the LWF detailed description at the link above, and it's quite obvious really. 32-bit formats, unlike 16- or 8-bit ones, contain enough information about an image for the reproduction of linear colour space, thus providing detailed correction of an image's colour and brightness. So there is some sense in making all the corrections before converting the image; stay in 32-bit mode as long as possible, and go into 8-bit at the end of work, or when there is a necessity to use software which doesn't work with 32-bit formats (most plug-ins for Photoshop or After FX).

In my case, I made some adjustments to the gamma, brightness and contrast with the Exposure tool, with subsequent colour corrections made using the curves. If it had been necessary to work with layers or use a lens flare in After FX, for example, then I would have converted the image to 8-bit (Fig.20a).

Fig. 20a

Fig. 20a

Fig. 20b

Fig. 20b

The post-production of night renders was different to the day ones, mostly with my active use of the lens flare to make the headlight reflections and glares on the hull more realistic (Fig.20b).

And that's all! I hope you have found this helpful. Cheers!

Makar Verigo

Fetching comments...

Post a comment