Before I do, I felt its worth also mentioning a fairly amusing comment I got from someone who spotted my work online. They thought I'd been posting up photographs - and said that I should really make them look 'less real and more 3D' otherwise nobody would realise I was creating 3D images...
Its not often I will hear someone tell me that they think it looks too real and ask me to make it look more computer generated. Too funny, but I know where they were coming from on an artistic direction.
 |
| "This is too real... Make it less real so its obvious that its fake" |
Its all in the software
For this challenge I decided to use LightWave rather then Maya. Originally I thought it could be a great challenge to pull this off with Maya (having taught Maya exclusively for the last 4 or so years) but knowing the amount of time that some areas of Maya's workflow take was enough to warrant going back to my roots (as it were).There are also some tools and features that are just not found in Maya, and these would allow me to get what I wanted at the quality I wanted with the tight time frames of a daily project.
What took the most time?
In the last few images, a series of isometric-styled computer hardware renders, production time and rendering can be done in 1 to 1.5 hours. That's because they are comprised of simplified iconic models with basic colour textures.
However for the first 20 or so, an average of 3 hours was normal with some of the more complex ones taking up to 5. These I attempted to go for a more complex realistic visual, which is why the time was so vastly different to the more recent projects.
In general, my gut feeling is usually 10% modeling and 90% everything else. In this case, modeling objects was more around 15%. Where the most effort seemed to go was on the composition, lighting and rendering. Surfacing didn't take as long (with fairly low amount of Photoshop effort required - but more about that later) for most things, however the amount of additional tweaking I would do while rendering was where this would increase.
 |
| Hmmm, deciding on an angle... Tricky... |
It takes a surprisingly long time to get that shot just right, and often the shot I had originally envisaged didn't look as great on screen once it was rendered - which meant a lot of adjusting and testing. However this is the area where LightWave really shined...
Don't over-model...
The tip of the day - just model what you see in shot. For instance in the first image I produced of a ZX81 I never modeled any of the sockets or insets of the machine because they weren't seen - and I simply positioned the cables to sit in the correct locations.Another example here is a BBC computer where the image rendered was of the BBC logo in the top-right corner. Did I need to model every key? Nup...
 |
| Model what's seen, not what you think you might see... |
Bumps, bumps, bumps...
A lot of 80's computers were built with plastics, and as we know, not all plastic surfaces are smooth and shiny. This is where adding bumps to the surface helps break up the surface and simulates the appearance of these types of materials - and I can honestly say that I did this with literally every project (given just how much plastic was involved obviously) |
| It may be tiny, but its essential to getting the look |
 |
| Bumps, bumps and more bumps... |
I also took advantage of LightWave's gradients to assist in controlling some of these to behave the way I needed them.
 |
| Using a slope-driven gradient let me produce smooth edged keys on the BBC |
Cameras
To pull off some shots, many of the scenes were fudged. That is, elements are moved, scaled and rotated to suit the shot and assist in making effects like fogs and depth of field behave the way that you need them to. Its pretty common in CG to 'work to the camera' with anything - its often the only way to pull off many shots. |
| Tiny tapes, HUGE TV set, fish eye lens and fog to haze it out in the distance |
I also took advantage of a Photoreal camera type that allowed me to apply a lens profile to the render to generate a true fish-eye effect. While that effect can be faked with Photoshop, you won't get quite the same result as a camera that sees the scene around it by processing the render through a lens.
Depth of field can also be viewed in real time (and very quickly) in the viewport, and that assisted a lot in fine tuning the final effect.
 |
| DOF preview in real time helped a LOT in getting the shot |
Themes
I knew I wanted interesting interesting ways to display items, and while camera angles and various rendering styles are one thing, I figured I'd also try playing with how to interpret the ideas. One in particular was when I wanted to produce something to represent the arcade game Dig Dug. Originally I'd thought all the video game related artwork would be sprites built from 3D cubes (which can look cool) but in this instance I decided to try something different.We'd been talking at work about post-it artwork just the day before this piece was done. In the last building we were, I'd drawn up some Defender sprites on the windows with post-it notes... And I figured why not replicate that same idea in 3D. Then I thought a glass window or office may be a little ambitious - so I went with what we do at work often - producing display boards and hanging the work from thin nylon... Well, virtually that is.
 |
| Virtual art exhibition, anybody? |
Lighting and Rendering
Along with surfacing, one of the key features that LightWave has is its Viewport Rendering - or VPR for short. This feature alone is worth its weight in gold, providing real-time rendered feedback as you adjust and move elements about in shot. When I was surfacing I used this a lot to see the result of tweaks to bump maps, reflections and node-based materials. As its a render, it also allowed me to preview my fish-eye camera, as well as depth of field and adjust as I needed.New Style
All that photo-real style work does take a lot of effort in modeling, surfacing and rendering. So, I decided after 20 days had passed I would swap my style and try something new and different to keep the work interesting... I also wanted to see if I could make things faster to create. |
| A whole new style |
Advantages here - I didn't have to worry about fine detail modeling, as I was after the shape and form of the machine and not an accurate model. Keyboards could be simple blocks, icons and branding didn't have to be added to the model... Though I decided to model the branding and place it next to the machine.
 |
| Taking a break from detail modeling and rendering for a while |
The shading was all done by driving the surface luminosity (or incandescence in other applications) through a mix of occlusion shader and a gradient ramp driven by the surface slope to allow shading of curved surfaces. The floor was a flat plane and simply used a radial gradient to create a 'spot light' style effect from the center.
That meant no need for lighting, calculating shadows - which then made render times a little faster to boot.
So yeh...
There's still 70+ days to go - its a fun project, though at times it feels like it overwhelms everything else around me when I know I have to complete something that night. With work, time is tight so I squeeze in things when the day is done and I'm sitting in the office...I'll do another update when I make any new changes in direction... And obviously a post when I'm complete! Feel free to check out the project so far HERE.
