- Is never satisfied with quick/shallow answers? - Wants to know the _why_ behind how things are built? - Gets lost down the rabbit hole searching for answers?
Then I hate to say it, but you have Rabbit Hole Syndrome.
Join me down the rabbit hole as I dig through different tech to find out how they really work, why they were built that way, and how we can use this to build better products in the world.
I just want to come by and say that you are probably the SINGLE BEST TEACHER on RU-vid for me. I think you are the best for my specific learning style (which is also detail-oriented/rabbit hole/curiosity driven). Thank you for taking the time to produce those amazing videos, you sure have a fan club of yours!
Sir, Thank you 🫡. I have difficulty focusing to videos like this and I watched many tutorials before and often got bored & ended up closing it. But your explanation was so good and I enjoy learning from you. Thank you for sharing your knowledge to us God bless you sir! I wished I discovered you earlier when I was still on college last year, my capstone project could've been different.
An excellent comparison of embedding tools! Another free option worth exploring is Hugging Face's Transformers, which offers robust embedding qualifications. #AIEmbeddings #HuggingFace
Incredibly well explained! I'm glad I found this as my introduction to geometry nodes. I'm still a total beginner with Blender, but I'm amazed at what it can do.
greg, this is an extremely valuable video, i find it really useful. you truly dive deep into the holes and present the findings so clearly and elegantly. as a severe RHS patient myself, i salute you, greg, thank you.
Some pointers with displacement, bump, and normal maps. Displacement: Can be height field (like a bump map) or a vector displacement map, which connects to the material output. So all shaders will be affected by it. You don't *have* to use true microdisplacement to use it (there is a setting for bump only), but I would recommend it; if the height field texture map is to be used as a bump map, you can use microdisplacement to find the ideal displacement height and use that as bump distance (bump effect always 1). Bump maps: Can only be a height field map, but the result is plugged into individual shaders instead of on the material level, so you can have a bumpy substrate with a smooth topcoat, impossible with displacement without additional model for the smooth part. Ideally you should only use Bump Strength 1 and adjust Bump Distance (or the height input) instead to obtain ground truth normals that reflects what a displacement would do. Adjusting Bump Strength below 1 tend to introduce some weird clipping of the input. Bump maps requires on the fly *CALCULATIONS* of the new normals, something a normal map provides by simple lookup. Because of that, it is scale dependent. More on that below. Bump maps require no UV layout and less prone to brake (i.e. box mapping), are way more flexible to mix and match, and using the microdisplacement trick (above) you can find out the distance to use or spot if there is any math functions that break down easily. Bump maps are best used for small details because stepping issues will be less visible, but does add computing time. Normal maps: These provide the normal offsets by simply looking up the color. However, to work properly these require a UV layout, are tricky as hell to mix and match (correctly), and can be prone to errors (box mapping). The angles in a normal map are *NOT* affected by texture density, as the color lookup remains the same. It's as if it's the height of the mountain was automatically reduced to maintain the angles, compared to the bump map where the height is maintained (producing steeper angles). Make a triangle wave and array it. Scale on all axis gives you what normal maps do. Scale on x axis only give you want bump maps do (without accounting for it, anyway). Normal maps are preferred by game engines, since angles don't have to be computed (by the bump map slopes) and thus impact performance less. They are also better used for large details like large chamfers since it's done with a single color indicating that angle instead of a gradient. Which means you typically get away with smaller maps and 8 bit (per channel) instead of large 16/32 bit floating point maps to handle gradients (if stepping is a problem) for bump maps.
Note that under object data properties there is a texture space tab. Using generated (and/or object) texture coordinates you can adjust the position and size of the bounding box that is used. If using XYZ tangents for anisotropic shading and it looks off from what you'd expect, chances are the texture space has been automatically updated to something less than ideal. Generated will use 0-1 in XYZ for the bounding box, whereas Object will use origo and allow negative coordinates in unclamped space. The old modulo math function was prone to breaking with negative numbers, so had to account for that, whereas the new (truncated?) fixes those issues. Generated also accounts for some modifiers like bending. Object coordinates are great for box mapped objects, but we have to adjust the coordinate orientation to change how the mapping aligns, as there are no way to rotate the coordinate system with nodes prior to be used for the image texture node.
Use case for strings I'm constantly using, is showing a bunch of information related to the object. I'm working with blended projectors and LCD/LED video walls, and some of the information I show are: Aspect ratio, overlap (pixels, projectors only), individual image resolution (per cabinet), total resolution (per video wall), calculated pixel density/size (to determine readability), lens shift and throw calculations (within physical limitations?), and what projector screen is being utilized (may vary with size due to production limitations). Snapshots of this is then used internally to discuss around the solution, and maybe sent to gfx to prepare pixel perfect content to be used in the final renderings (although I use UV vector snap to force anything pixel perfect anyway). Sometimes things are used for communication purposes rather than rendering. One problem is since having to use nodes for separators, concatenations, and special characters, it can get really messy really fast, so all this is dropped into a separate node group. I wish there was a more elegant way of doing this using a multiline expression interpreter.
Thank you for that video. Learned a lot from it. Though what I was really hoping was that you'd show how to make a node that would stitch around a shape, rather than just being a circle
13:00 hey, might be bit late, but wanted to point out that you can also unstage selected ranges in vscode, so you could just stage the full file and unstage things you wanted to keep for next commit
i'm a beginner in 3D art and design world, i've chose Blender, so i'm a beginner in Geo nodes and these information are very useful to me, i git the concept, thanks sooooooooo much
@ the end: I'm not certain, but I would assume material index refers to material slots per object. You can have multiple materials on separate parts of a single object using material slots. One thing that would make sense is to put the material selection as a group input, so the user can add whichever material they want right in the modifier.
Thank you, Mr. Professor!╰(*°▽°*)╯ I've watched many videos, and each of them varies in style. I couldn't replicate what I did because I kind of mixed what I learned from them, which confused me. This opened a window that I was passing by every time. I didn't learn basketball by imitating Michael Jordan; I realized at a later age that the key was to understand and master the fundamentals. Thank you for sharing the fundamentals. I'm subscribed. 😍
