I didn’t know how to put this, because I’ve seen many techniques for it, and none of them I’ve found are ideal for a CAD to finished product rendering workflows.
My current workflow, not in blender, is to get graphics in actual size in illustrator. They are black and white images representing each printed color on a product, shown prependicular to the tangent of whatever the graphic is (usually that’s just straight on front, side, top, etc, but sometimes will be at the angle of a curved surface, which gets trickier). I bring that graphic in, and it’s already at the correct scale, defined as a plane, and place it. Using that black and white map, I can just use two materials. One is an instance of the material that is everywhere else, and the other is the that color of the graphic. If it’s multiple colors, you get multiple black and white maps, and need to layer them properly, just like they do in manufacturing.
So in blender, it looks like lots of people are using a shrink wrap method. Simplified, you add as a reference image plane, which adds the image at the correct aspect ratio. Subdivide the heck out of it, place it near where you want, and shrink wrap it to the other object, with a tiny offset. But, of course, the topology won’t match, so you have to go Crazy with subdivisions to not notice, and use a careful balance of offset, but not so much you can see it. Also, the transparent areas of that plane cause issues with denoising, so get ready for a pretty high render time.
I can add dummy and place an object as a separate UV channel, but have the added issue of projecting all the way through the object, and if you have a specific size, that gets Really difficult to control. Like, the aspect ratio? I’ve got to do math on each one? Sure, if it’s my own workflow I can stick to 1:1, 1:2, 1:4 ratios to be sure the math is incredibly easy. But often times I’ll get someone else’s work and it’ll be something like .223in x 1.47in at 455dpi. Yay for very specific manufacturing specifications! And yes, that has decimals in the pixel size, and no, that’s not ideal (but a half pixel off is better than the tolerance of manufacturing, so I’m ok with that.
I ended up purchasing the Eyek add-on. https://gumroad.com/l/eyek_blender_addon It’s really cool, but really it’s just baking the textures to a map. You have to create a Really high resolution map, AND have to actually have your object decently unwrapped. CAD data imported does Not unwrap well. There are Tons of vertex problems that you don’t really see until you modify Anything at all, because it has custom normals.
Does anyone have a decent workflow for adding labels? And for those that think a method involving a couple extra steps for each label isn’t a big deal, imagine having to label a keyboard, where you have the letters as one map, the function keys as another, a second color and second graphic next to the function keys, the media keys, and the special characters above the numbers are also a different color. After you get it all set up, the client comes back and says “we’ve decided to change all the keys from black to a slate navy. And also we added a dished shape to the key to make it more comfortable, so update the geometry as well.”
Now, suddenly, doing this with just maps is super helpful. But setting up the maps isn’t all that easy.
Ideally, I’d like to do things Just like Eyek, where I import a texture as an image plane, and place it where it needs to go. But then just use that plane to position the graphic on it’s own UV map. Where EyeK is actually baking that to an unwrapped uvmap, which loses a LOT of quality if something like a keyboard needs to be a single object.
And an extra bonus to you if you also know a good way to create some sort of masking technique that allows you to specify a depth from the plane so that it doesn’t project all the way through an object (but without having to uvunwrap the object). I’m sure it would have to be through an add-on or something, but it could use an extrusion from the rectangle at whatever depth, then only project onto the faces that intersect that box.
Anyone have a better technique, that is as procedural as possible?