Graphics Creation Test
It has been a while since the last progress post. I was busy comparing graphics creation approaches for Nordenfelt. I tested 2D pixel art and 3D modeling in a performance confrontation.
Spriting approach:
For testing this approach I created this basic draft of a Steampunk airship:
First I fleshed it out with spriting techniques and measured the time I needed for each step:
- adding details like pipes, cockpit glass, etc.: 1,5h
- shading: 3,25h
- coloring outlines: 0,75h
- coloring body: 0,5h
This is the result:
The whole process took 6 hours of work. There is still room for improvements like texturing or holes which show some machine interior. But it serves the comparison needs so I stopped at this point.
3D modeling approach:
I used the same 2D draft for creating a 3D model from it. As with the sprite I timed each step:
- modeling floor plate and body: 1,25h
- modeling cockpit glass: 1h
- modeling rotors: 0,25h
- modeling tanks: 0,5h
- modeling pipes: 0,75h
- modeling rivets and joints: 2,25h
This 3D model came out:
For the final bird's-eye view image materials and rendering had to be adjusted:
- setting materials: 0,5h
- setting lights: 0,25h
- rendering/post-production: 0,75h
After 7,5 hours the result looked like this:
Comparison:
The advantage of pixel art is the speed how fast you come from draft to the final sprite. This is due to its low resolution which prevents you from adding unnecessary details. The smallest entity is the pixel. You can't get smaller.
3D modeling needs more time from draft to the final rendering. But the advantages outweigh this disadvantage. Scalability, rotatability and easy experimentation with lights are much easier when you use 3D models for 2D sprites. Pixel art is impossible to scale or rotate without ugly distortions. So you have to manually create a frame for each scaling or rotation. Pixel lightning is purely handcrafted and a pain if you can't figure out the lights right from the beginning.
Finally animation is a big plus when 3D models are used. Pixel art needs a separate sprite for each animation frame, the same problem as we have for scaling or rotation.
Conclusion:
Disillusion.
This test showed me that pixel art is hardly applicable. For preventing detail bloat in 2D sprites it is necessary to keep their sizes small. This results in smaller resolutions and a smaller play field. Only retro fans would be pleased about a low-res graphics style. This would shrink the potential market too much. Therefore I have to come up with a new way to produce 2D graphics which allow a higher resolution and better production scalability. I'm going to think about this a while.
If you know other 2D graphics approaches: Please tell me!
Cheers,
Thomas
Comments
I would definitely agree that if you are adding in light shading seperately, time would be eaten up rather rapidly for Pixel artwork - many of the other points you make are quite valid.
A few considerations, on this topic:
* Expense/requirement for 3D modelling software, and experience in using it
* Semi-reusability of pixel based spritework (You don't start from scratch with each frame, same as with cartoon animation).
On the whole, I would agree with you that for more polished graphics, 3D modelling would be suitable. However, I would also argue that 3D modelling allows a little too much potential for feature creep (ie: Oh, and I might add a gun here), as well as requiring time to process the image from the model -> 2d sprite image.
Man, that feature creep! Adding 3D details which never will be visible in the 2D sprites. Stern discipline required!
I had considered Blender when I made the comparison. That said, as someone who's more familiar with industrial packages (yay, Solidworks), I'd definitely be more pleased to have those available (and boy, those cost).
I think the main issue at heart is twofold:
* What skills does your artist *already possess*?
* What look/feel do you want?
Any look/feel is acceptable (even to the level of Ffffffffffffuuuuuuuuuuu style comics drawn in mspaint), if the entire game is presented in a consistent style.
When we try and mix 2d with 3d, it inevitably causes the following:
* Inconsistent look/feel
* Significant amount of manufacture time (make 3d model, convert to 2d image, rename files, adjust for detail level required, clean up jagginess)
Not sure if you're aware of this, but there's also some discussion of this post over at reddit.com/r/gamedev: www.blackgolem.com/.../109-graphics-creation-test.html
www.reddit.com/.../comparing_creation_of_pixel_art_vs_rendered_3d
I did this asset creation test to find out which approach I should use for my game: www.nordenfelt-thegame.com. I've seen the discussion on reddit.com and agree with most of the statements. The article wasn't meant to be general wisdom but personal experience. I should have mentioned that somewhere. Finally I'm happy that this article triggered some interest.
Thanks for your comments. Your blog is cool, especially the pixel art. I like it!
modeling cockpit glass: 1h
modeling rotors: 0,25h
modeling tanks: 0,5h
modeling pipes: 0,75h
modeling rivets and joints: 2,25h"
Actually my guess is that the time required can be significantly shortened with some practice
Although it largely depends on the complexity of the model
There is also vectors with vector shading with tools like inscape
Thanks for the Inkscape suggestion. This tool seems like a gain for my art tool collection. It brings new opportunities to my asset creation habits.
RSS feed for comments to this post.