The Global ARCHICAD Community

Stay informed. Get help. Share your knowledge.

View only topic: Collection of Best Practices edited by the moderator. (Please post your tips into one of the topics under section II.)

Moderators: Karl Ottenstein, LaszloNagy, ejrolon, Barry Kelly, gkmethy

User avatar
By Dwight
#25238
A picture is just a bunch of dots.
Dots of texture pixels.
Dots of color as rendered pixels.
Dots of color perceived by the eye and re-rendered in the eternal cycle of image gathering and reviewing by the internal workings of the brain that finally understand the dots as an image.
Everything else is just colored bubbles.

Think of making a rendering as looking through a screen or picket fence at reality – you only see part of the information, but your brain finishes the job of interpretation. If you look closely, you’ll see that only the center of the field of view is sharp – your eyes dart around and your irises compensate for light changes to help the brain form a memory of a space that is uniform – totally unlike the reality of glare and darkness found in a real situation.

Success with textures is found in understanding how the brain perceives these dots – still or moving, and how the computer merges adjacent dots into what appear to be smooth surfaces or edges.

When you ask how big to make a texture, that is a big question and I am afraid that I will have no time left to answer anyone else after it.

If a texture is to be smooth in the rendering, it should be a quality image to start with – no lousy JPEG with its rosy fringing. I have some 3Mg wood grain textures used for close-ups. They don’t “break” at any scale, but if the whole image was filled with textures this size, rendering time would be eternal.

DEPENDING ON THE SIZE OF YOUR FINAL OUTPUT, your smaller textures will fail – becoming smeary and videoed, where the component RGB colors show up, especially where the camera gets too close – any situation where the size of a texture pixel is bigger than a rendered pixel. While it is more difficult in the LightWorks engine, the ArchiCAD engine tells you exactly how big a pixel of texture is. A one hundred pixel square texture made to 1 metre yields texture pixels a centimeter wide…. And if you approach closer that this in your rendering, ArchiCAD’s engine tries to make up the shortfall in information with anti-aliasing - selective blurring/smearing of data – sort of like putting pots in the oven when Karl Ottenstein and his wife show up unexpectedly for dinner. “Hide the mess! We have company!!” There is a slider for anti-aliasing in the ArchiCAD rendering engine. It adjusts from “nominal,” to “for ever.” I’ve just had a terrible experience where a granite counter top was not aliased enough. Looked fine in the still test but all lively and scumbling around in the animation.

That assumes smooth textures. The inverse is patterning, like a checkerboard, where you want pixel edges to show and be sharp, In this case, the smaller the better, so that each pixel gets huge and its rectangular edges show. When Abvent first brought out AV_WORKS, it blurred all edges and it was impossible to get good edges with ArchiCAD materials…..

No ArchiCAD operator should delve into texture issues without a photo editor capable of layers, filtering, cloning and alpha channel creation. And if you haven’t heard about alpha channel you should do formal research into this invisible data channel that controls special reflection, transparency and bump mapping aspects of texture behavior.

You can easily create that cedar siding with bevel – it is an exercise in my seminar that many have enjoyed. Take a noisy color texture [how weathered IS the siding?] equal to the height of two siding boards – say 100 pixels square, and draw your shadow on it with alpha channel information. It is a grey scale channel where you draw your effect – a shadow definition, say. You’ll do two things: define the shadow under the bevel by making a full strength horizontal line equal to the depth of the bevel you want, blurring the top edge against the nose of the siding board one pixel and the bottom edge three pixels to soften it. It looks like a shadow, cast from a board.
Then make a subtle gradient downward from the shadow stopping just before the nose of the siding board – leave five pixels untouched. This emulates the subtle change of sky reflection made by a tilted board and results in a highlite at the board nose.

When this texture is properly sized to equal the real siding size and used in a texture with BUMP MAPPING turned on in the material, the shadow created will define the siding edges and will disappear in shade, just like siding shadows would.

Is this bevel or lap siding? OOps. The principle is the same - using shadows defined with bump maps to suggest the profile of a uniformly colored element....
User avatar
By mallehk
#25239
Thank you Dwight for that very thoroughly answer. I can see that I haft to go deeper in to Photoshops capabilities (alpha channel), for making my ceder siding. Please notify me if you get to Denmark, a lecture in rendering would be great.!!

Greetings Martin
User avatar
By Dwight
#25240
And it reverts to the old truths about drawing - how to render an object of uniform color - doesn't help to put a black line around things - that is a "cheap draftman's trick," according to my beloved mentor Gustavo DaRosa - a man unmistakeably having an architect's name. Imagine :
"Hello, I'm Gustavo DaRosa. Would you like fries with that?" Doesn't quite work.

In fact, in the computer age linework of any kind used to define surfaces is such an old concept that they called it a draftsman's trick, not a drafterpersonnaginess's trick like they would in this politcially correct milennium.

So much is achieved with the alpha channel and it is no more work than drawing with color.