crack help

davidmajdi

hi guys / i don't know why but my crack in SD is not good / i mean the depth and random thikness are not aceptable
if yoy have any tips or video link plz suggest me
this material i did last week but i'm still dont like cracks/ these are not normal i think

m_asher

I've had some decent luck using Flood Fill to Gradient and Vector Morph to shift the cracked pieces slightly. It keeps the cracks from feeling too uniform by pushing pieces into the grout space between tiles, and it's easy to give the cracked pieces a gnarlier edge with the Bevel node + Slope blur.

Clark Coots

One of my Quick TipS could be helpful for you: https://www.artstation.com/artwork/mqWQ9d

davidmajdi

Clark Coots said:

One of my Quick TipS could be helpful for you: https://www.artstation.com/artwork/mqWQ9d

thank you

gnoop

I believe  doing cracks proceduraly is a waste of time .   A kind of sport for a sake of sport.    It's still  that cell noise  originating  from blobs intersecting,   instantly recognizable   whatever shifting or deforming you do.  No proper physically driven branching.   You can waste a week and make it a bit better looking   still you wouldn't deceive  anyone.       

I usually just take some photo of cracks done by a phone , whatever blurry it may be or just hand painted sketch   and process it through edge detect, flood fill, a bit of extra warping etc.     

poopipe

That's the pragmatic approach and I'd urge anyone without a decent understanding of maths and procedural generation to do as gnoop says if they want plausible results. 

I've been working on and off for years on ways to generate a believable and robust system for cracks and I've yet to arrive at a solution that isn't basically shit. 
Recently I've started working on growing tree like distance field structures in designer, this seems like it might be the way to go but it's complex, expensive and (based on watching people use my other sdf based tools) unintuitive. 

Clark Coots

I disagree doing cracks proceduraly is a waste of time. You make good points, it's mostly cell noise, tends to be recognizable, no proper physically driven branching.

However, if it's procedural you still get the flexibility of different sizes, widths, seeds, and can be re-used on different graphs with different results. You don't have to rely on a photo dependency, everything is contained in your graph. I'd also argue even with a photo sourced cracks, it may not be physically correct depending on the material you are using it for. Ceramic tile cracks different than concrete, etc. Dpending on your under laying height, material type, history of weathering, age the physics of how it cracks in the real world would not be correct unless your are 100% photogrammetry or possibly simulation software. Anyway you slice it cracks and materials are artist interpreted and will never be "correct", just has to be believable.

That all being said I'm all for alternative techniques and more accurate solutions like L systems to help create different and hopefully better results.

I personally think these cracks by Daniel Thiger are fantastic and showcase the potential and the variety you can get with procedural cracks.

poopipe

That's a curated selection - not to say some of them don't look pretty good but it's not a general solution in that you can't apply a random seed and reliably get a plausible result - you're basically left changing numbers until you get lucky. 

Perlin, voronoi, poisson and most other forms of cg noise are built around the distance between a field of semi randomly distributed points which is kind of the opposite of what you want for cracks - which grow outward from points of stress 

Clark Coots

Agree, there are a lot of happy accidents with procedural things, not all combinations of variables give desired results. I still stand by that it's not a waste of time to do procedural cracks. I think in the age of Designer we are seeing some of the most interesting and highest fidelity materials being created. As I said I'm all for alternative methods that produce better results, perhaps I'll do some investigating

gnoop

in this example Clark Coots posted  the  blobs nature of cracks have  just been clever hided  by some other  straight pattern.   But once you need to do just asphalt  with some cracks , without dirt  and other hiding  patterns  you  would instantly come to a problem .   

Besides usually  you  can't put such style  of a texture in a real scene because it would  repeat so strong  , so ugly.    It's usually decals only that are free to be so distinctive.  

Regular tiling textures are usually noting like this.     In racing genre I am working for at least.      Any cracks  should be more generic   and it's instantly deprive you form  such  hiding tricks  used here.