Topic: Texturing Strategies for Blender 249

[ACDmvmkr]

In the tutorials, it mentions efficiency in applying textures to objects, namely, less is better - using the smallest image files makes the computer have to work less in generating the finished product.

My question is this - in making a large complicated object, with many parts requiring the application of a texture image to many different surfaces, is it better to create and use one large JPEG image (or any other type of image file), and then map the various parts to different portions of that one big image, or is it better to create many smaller individual images for each and every individual part?  (see below)


Multiple samplings taken off of just one image....

Part 2 of the above question - if you are sampling just one portion of an overall master image to texture a part, is it better to map that part to only a small portion of the image, or to let the map for that part spread out over a larger portion of that image?

For example: if I'm providing color to a small part using a portion of a large master image that displays that color, does it matter what size I allow the unwrapping of that part to be on that image (as long as either way, the map size only covers the desired color without crossing any unwanted details elsewhere on the image) that I want to apply?

(see below)


smaller mapping of part


versus the larger mapping of the same part

Which is the more efficient?

I don't know the mechanics of how the computer imports these image files for use in Blender, or which way is considered more efficient. When Blender applies texture images to all the parts of a project, does it separately import an image file each time it applies to a part, or does it import once, and then re-use that image each time it is called upon for various different parts.

I hope I got my point across, however clumsily....
Thanks

[kat]

There isn't a simple answer to this one because it depends on the object, what it's being used for and what rendering tech is being used. For the latter, there are often hard coded limits which determine what you can safely use without causing performance issues, so using larger textures because you can isn't always the best approach.

With regards to the object and how you UVW unwrap that, your defining criteria will likely be texture density - load up a checker texture, if that appears the same size over the entire mesh then you're on the right foot. If it varies, or you can't get the entire UVW map into the available space then you've got problems which will likely mean having to split the mesh into sections and over several texture assets - it's important to understand that it's not necessarily about texture size, but rather texture usage.

What you want to be avoiding doing is making UVW segments so small that they become 'pixel sized', i.e. you're using just a single pixel of a texture to map to a given UVW, with isolated elements that's problematic as it can be noticeable in game if that element is sitting on or near something with better texture application (density), the difference can be striking. If that 'pixel' is part of a larger object then it's not normally an issue.

Having said that, you can get away with doing that if, as you ask, your mapping a solid colour to the element; so long as you leave enough border around the UVW island it shouldn't matter how much or how big the UVW is in relation to the texture.

You also need to avoid doing what you've got there where you have many, many small elements all individually mapped.. if those are separate segments within the overall mesh try and find a better way to UVW map them so if the object isn't necessarily going to be 'unique' in nature, chop it up and mirror/duplicate sections so you're using the same UVW mapping over several of the same elements.