art with code

2016-12-08

Acceleration Design Concepts

Continuing from the Project Proposal, here are some design concepts and notes that I jotted down in my phone notebook (Samsung Note, it's great. The last couple ones are Photoshop.) Some of these are a bit cryptic, especially the last page, which is a scattering of random ideas, let me do a quick briefing.

The core idea running through the design is avoiding having to model by typing in numbers. The layout would be based on snapping and grids to get clean alignments and consistent rhythms. For example, the default translation/scaling/rotation mode would be snapped to a grid. That way you can quickly block things out in a consistent fashion, and go off-grid later when the basic composition is solid.

Another thing (that's not shown here) to speed up creating compositions would be repetition, randomization and symmetry tools. Throw a model into an array cloner, pick the shape of the array, tweak randomizer parameters for the array, set it symmetric along an axis: very little work and you get a symmetrical complex model. Add in a physics engine, and you can throw in a bunch of objects, clone, run physics and get something natural looking very quickly.

As the concept behind the app is doing quick dailies, the default setup should already look good. A nice customizable skybox, animated clouds, good materials, and a classy intro camera pan. The camera would be a based on real cameras in that you'd have aperture size, focal length and depth of field that work like you'd expect them to. The exposure would stay static over changes to camera params, you'd have an exposure slider to adjust it. The camera would have selectable "film stocks" to change the color tone, and a post-pro glows, flares and vignetting.

I was thinking of basing the workflow around kit-bashing. You'd have a library of commonly used objects and materials (e.g. landscapes, rocks, clouds, fire, water, smoke, wood, plants, metals and so on) and could drag them from the library to the scene and build something interesting-looking and polished very quickly. The inspiration for this are UE4 speed modeling videos like this.



The tool wouldn't have modeling features, but focus on importing pre-made models and mashing them together. This would make the tool (that's already sounding pretty intense) simpler to make. The focus for the tool in my mind is quickly building and animating WebGL scenes in a WYSIWYG fashion: you'd always see the final render quality and performance, and could work with that (instead of having to guess).

No comments:

Blog Archive