Hello guys, would love to hear your thoughts on how to model the handle of the scissor sword from the anime "Kill la Kill". I've had the rest of the sword modeled for ages but I just can't wrap my head around the curved part of the handle. I'd be happy to use dynamesh even, as long as I had a close enough lowpoly done in max first. If you want more refs directly from the anime check here too (shape changes slightly, i'd focus on the prop). Any feedback is greatly appreciated!
Yogensya: As always.. Show us your own steps and approach to figuring this shape out. It would be a disservice to your development if someone just flat out showed you exactly how to go about it.
Basically the shape is just a extruded diamond where you take half of some of the cross sections and move them a bit to the side. It's very simple and I strongly recommend you figure it out yourself. Keep in mind this is much more about being able to see the shape than being able to model it, so trying to paint wireframes by hand on the photo might help you.
This is the closest I ever managed to get to the anime references. The cross sections were almost there but the round tip is another conflict area for me. Looking at it now I'm not sure why I didn't use a bend modifier on it. Modeling the prop version I showed earlier seems even more confusing to me though, at least if trying to use subdiv modeling..
The topology is really bad though on this mesh. Every time I look back at the project and try to approach this piece I get a block. Either way I'll see if I can rethink the whole piece a bit and start fresh.
Yog: Well, you should try to do the wireframe overpaint I suggested. Now the shape you are modeling is not at all the same as the one in your reference. There's just a vague semblance, as if you're using your memory of the picture as reference as opposed to the picture itself.
You can set the reference photo as background image in the 3ds Max viewport, apply a transparent material to your mesh, and get a perfect match by placing one on top of the other.
Concentrate on simply modeling without subdivision to nail the nature of the shapes and flow. Once you've successfully done that, apply subdivision and get your hard edges down.
At the moment you're trying to do both things at once, and failing at both. That's unnecessarily complex. Like, would you peel potatoes while boiling them? Peel half the potato, stick that half into boiling water while simultaneously peeling the other half. In theory that might sound more efficient, but in practice it would be a nightmare. Email me for more potato allegories.
@Yogensya It's like you're trying to make mashed potatoes by chewing a raw potato and throwing yourself in boiling water ^ some shape id I might do in my process.
^ I would then start "sketching" each shape while planning on stitching them together
^ Sewing these shapes and adding dimension
^ Mirroring the handle and sewing - You can then bevel or crease (or quadchamfer in max) - bare in mind I was not that careful - you'rs should look better if you take the time.
I have quite new to modeling and have been trying to model this cap from this water bottle. Attached is what I am trying to model and my awful attempt. Really running into a wall trying to model this, any advice would be greatly appreciated.
@rogue1 - Not exactly the same proportions, but something like this topology would work. I started with a 1/8 cylinder. then I made a part of one side flat as on the reference, made the curve . Then I just duplicated and mirrored the slice around.
You'd just select some of the verts of that horizontal edge that goes across the middle of the cylinder and you move them down slightly to make a curve.
@Obscura I played around with moving some of the verts on the flat edge down to make a curve but I am stumped on how to get the indent that doesn't mess up the vertical straightness seen in the reference image. Sorry for my lack of knowledge, learning :P Thanks so much for helping me!!
Oh, forgot to say - the shape is a cylinder-box intersection. A quick boolean is all that's needed. Also two of the three edges on the flat area can be repurposed to be control loops.
booleans are the devil! patch modeling 4 life! lemme alone perna you know I'm lazy - I dynamesh everything anyway :P (mostly want to hammer home the early evaluation steps ofc)
Proportions might still not be correct, I was not really going for that! Adjust segmentation for your needs. ...
Im showing a sequence, steps after each other... - On the first one, we can see a 64 sided cylinder - other segmentations would probably work too. - On the second one we can see a 1/8th of a cylinder - since the pattern in very symmetrical... - On the third one we can see that 1/8th cylinder with an edge in the middle... that will serve the base of the detail... proportions are incorrect, adjust height and segmentation for your needs! - on the fourth one we can see that edge got chamfered - on the 5th one we can see that the edges got adjusted, i welded some of the bottom vertes to their top neightbour, then I adjusted the remained curve - on the 6th one, I pushed back the inside verts of the curve shape - very top verts and the ones below that - so I got that flat surface - on the 7th one, we see no change, I'm showing the shape from step 6 - this should make step 6 easier to understand and implement - on 8, its just symmetry, copy, rotate, symmetry... - on 9, I started adding support loops. - 9, add them inside the curved face - 10, adjust the support loops that you made inside, so they evenly follow the supported edge
I wont show the top part, that should be obvious
You should get this result after applying subdivision:
Hi, I've got problems with this mecanum wheel. My approach is quite clunky, it's two separate pieces, one bended and second cloned with array, and after that both turbosmoothed, and finally merged, welded and turbosmoothed again. After all there's some flaws. Is there a better and simpler workaround?
Hi, I've got problems with
this mecanum wheel. My approach is quite clunky, it's two separate
pieces, one bended and second cloned with array, and after that both
turbosmoothed, and finally merged, welded and turbosmoothed again. After
all there's some flaws. Is there a better and simpler workaround?
Step 1 make a cylinder with 40 sides or more.
Step 2 the rest ... duplicate rotate with pivot at center ... do the rest
Used this thing to help me have quick smooth (sneaky tease lol)
@perna I Honestly don't get it , This image is right lol ? If you talk about the pinching that can be bought by edges near each other in the depth direction of the Cylinder I second that , Smooth good on my side anyway
@perna I Honestly don't get it , This image is right lol ? If you talk about the pinching that can be bought by edges near each other in the depth direction of the Cylinder I second that , Smooth good on my side anyway
The construction of the mesh is broken and the shading is not good with all those messy triangles and ngons.
If it wasn't clear, my advice was to twist the entire cylinder. That way the intersection lines up perfectly with the underlying structure, instead of being completely at odds with it like in the models on this page. This not only completely avoids any of the aforementioned problems, but is significantly faster to construct.
To make it absolutely clear:
Blue/Red = Other models. Messy, requires lots of manual cleanup and dense geo, leads to bad shading. Green/Yellow = Clean, logical and efficient.
GhostDetector: There's a lot of discrepancies between my mesh and the target design, none of which have any relevance to the point being made.
Below. Exactly the same mesh but with the design you mentioned. Makes no difference. To make this change required me to move two edges, everything else is automated. An enormous advantage of keeping it simple and clean. No manual cleanup work. Your mesh doesn't have control loops, btw.
The construction of the mesh is broken and the shading is not good with all those messy triangles and ngons.
If
it wasn't clear, my advice was to twist the entire cylinder. That way
the intersection lines up perfectly with the underlying structure,
instead of being completely at odds with it like in the models on this
page. This not only completely avoids any of the aforementioned
problems, but is significantly faster to construct.
To make it absolutely clear:
Blue/Red = Other models. Messy, requires lots of manual cleanup and dense geo, leads to bad shading. Green/Yellow = Clean, logical and efficient.
Will also point out that even If one would assume It's just an example, It's just not accuracy , In this context we have curvature and a hole ... In your just straight cubes , I know you can make it clean , the point is result on my part , and the fact I don't like to stress U (horizontal direction) and V neither (Vertical) , I assume you would accommodate all these new factors with edges running in the U direction . . .
"Will also point out that even If one would assume It's just an example, It's just not accuracy , In this context we have curvature and a hole ... In your just straight cubes , I know you can make it clean , the point is result on my part , and the fact I don't like to stress U (horizontal direction) and V neither (Vertical) , I assume you would accommodate all these new factors with edges running in the U direction . . . "
Dude...your mesh is just messy. Accept it. No shame on it...Beeing defensive and argumentative about a method that was JUST PROVED that is less efficient\clean is just....what....i cant even...
If i had to edit that mesh i would tell you to redo it in a way that makes edition on my end easy and not a pain that would require me to redo it...
So you guys are telling me you're getting High poly from others ? Even if the piece is "complete" didn't know about this... I just got it out fast as I can and in My way of working Pena's solution while It's the "cleanest" is not the intuitive thing for me. I really care about my surface on the other hand and there's no whoble
One of the main points of subD modelling is to have a manageable control mesh you can re-address and make changes to. If you are an outsource artist submitting these control meshes are nearly worthless if they aren't clean to make alterations to. In production you pass off your meshes all the time.
So fast and messy = time for someone else to clean up.
@perna Sorry, I posted that before I saw your explanation. (forgot to reload the page) I'm using blender and use a bevel modifier, so I use that as my control loops and clean up afterwards.
None of you are right, not even me, if you guys look again at the original picture that is made with CAD software, the chamfers are very big and intersect each other very nice without smoothing errors, to me that is hard to do with traditional sub-d.
If it were me i would do it in fusion 360 or even zbrush and call it done.
what if your client/studio doesn't use fusion 360 and wants a subD model ? - how would you do it in zbrush? What if you got it done and the client wanted changes? suddenly your $70 per hour job turns into $10 dollars an hour.
@s1dK ; maybe you're right but I'm not even sure OP will have had those hard chamfers, and honestly I wouldn't say no, to that in SUBD, yeah F360 and Zbrush are more efficient in the case and in most cases with hard surface though
WaYWO, the thing about defensiveness is it's more likely to harm your progress than help it. Better to just say "Thanks, I learned something today", and know that you're one step closer to being awesome at 3D.
Your mesh would never pass approval here. That might make you upset to hear, and you might feel that I'm a jerk or out to get you, unfair, unholy, sacrilegious or stupid, but it's fact, and one you could benefit from hearing. It's only workflow; you're not married to it.
Shape intersection is a core skill of subdivision and should rely on good practices and experience, not "intuition".
s1dK: No. A base principle of subdivision meshes is to alter the design to align with the technology. Also I can't even begin to understand why you'd suggest ZBrush.
s1dK: No. A base principle of subdivision meshes is to alter the design
to align with the technology. Also I can't even begin to understand why
you'd suggest ZBrush.
Because It's fast.
Your mesh would never pass approval here
ok but approval of what , be precise ? Some HP for low po Bakes ? I don't think you're hostile but Are you saying there's "a way" of modeling? sure nice topology exist, but there's certainly no " way of modeling " for me, and I'm not *cough* the first to say that , And I'm not even a fanatic of a particular workflow, I just do things that bakes right (no crappy surfaces from HP), please guys don't slice my throat I'm being honest.
we are just pointing that having easily editable/clean objects with nice subd flow is a good workflow practise. if someone else had to work from that it would be a nightmare and that person would be cursing the whole time.Good practises are also ones that dont make life harder for other artists that pick your stuff to work from.
we understand that it "works" with you , but that would never be approved by me if an artist sent me that .
it was pointed out what could be improved , now what you want to do with that info is up to you , but when you try to convince others of a broken workflow it defeats the purpose of this thread.
@Joao Sapiro Your work is inspiring my tools. But you want to know something funny ? I didn't even say @pena 's was bad I actually agreed if you read carefully . broken workflow is not the word you want to use though.
You're absolutely right about that, but once you go ZBrush you can't go back. Iteration on a mesh you've put into ZBrush becomes damn near impossible and could potentially cost you a lot of precious time.
Where'd the quote button get to now? TBH this assembly could be done with floaters just as easily.
@WayWO It's also important people learn how to make clean subd, and your support loops are anything but in this example. While the others would push for approval on their teams, this is the sort of thing I might buy off a website, and it would be a hard sell with that geometry if I knew I would need to change it to suit whatever I've got in mind. I know you didn't come to this thread with this kind of scrutiny in mind, but hey.
@TeriyakiStyle If the client/studio wants exactly the same mesh as in the concept you can talk with them and say` hey...with fusion we can get that exact mesh and trust me if you want to modifiy that mesh after some feedback it`s very easy compare to max sub-d.
Don`t get me rong, i use sub-d every day but for some meshes that has all kind of intersections like chamfers on top of chamfers it`s best to go with CAD software.
@perna i understand you but if your client wants that exact model you can`t alter the design just becase it`s not possible with sub-d, you must find another solutions to make that exact mesh, there is when CAD software gets prio vs sub-d.
Piggybacking the above discussion because I have a more general question about shape design that pertains to to it (somewhat).
I wonder if the parts below are made in CAD? The combination of very clean edges (non-beveled) with very complex shapes and intersecting model parts seem to suggest a non-subd approach? Or maybe just booleans (although I have to admit that even with a boolean approach in mind, I'm somewhat stumped as how to approach some of it)? It's a garage kit model and I'm pretty sure Netfabb was used in the end part of the process (to cut parts and make molds? I don't really know much about Netfabb), based on some other images from the same person.
I don't know, maybe I'm just a lazy person, but I'm legitimately overwhelmed thinking about how to approach this in subd...
Waywo: "yeah @gfelton ; most importantly people come here to learn Hard surface SUBD Modeling though"
I can't seem to quote individual posts so this will have to do..
I just wanted to chime in on the previous topic.
While it's true that you managed to accomplish the shape, the main problem here is not the fact that your mesh is all messy and shit. The main problem here is that you have this notion that, as long as you get the shape done it doesn't matter how messy it looks.
In some cases this is true, but with all aspects in life, we should strive to be elegant in our solutions and always be learning more and better ways to accomplish tasks in an elegant, clean and fast way.
This applies to everything in life.
There is always a smarter, faster, easier and elegant way to do things. Programmers will tell you this. There are coders who get shit done with their messy code with no regards to how clean their code looks... but then if you ask someone to maintain their code.. jesus fuck, what a clusterfuck. Do you mean what I'm saying?
Whilst Perna's post didn't show to to make the shape 1:1, it did show the most elegant solution to accomplish that in a clean procedural way. You can pick up that method and apply it to make the shape the OP wanted.
This is what learning is all about. Trying to figure out the best/smart/easy to maintain solution for if someone asks us to change something, you can do it easily.
I will give you an example it stuck with me when I was in high school. My math teacher once was telling a story when he went to the supermarket to buy bottled water.
The cashier was just a kid, and when he was counting the bottled waters to insert the number on the machine, he counted them one by one... My teacher was livid. He said, this is why you need to learn the basics and think logically so you put the least effort on yourself. In this case, the cashier accomplished his duty.. he managed to learn how many bottles my teacher was buying.. but he counted them one by one. Sure it worked.. but you know.. he could have just counted the vertical and horizontal number of bottles and then multiply them to get the same result, which would have been easier and would have demonstrated he knew basic logic.
Replies
Chop the two ends of the capsule off and all you have is a tube.
Model the tube flat with the extrusions, then turbosmooth it once.
Then bend it 360 degrees and add the two spherical capsule ends onto it. Then turbosmooth and it will work fine.
If you're having trouble making something subdivide on a spherical/cylindrical surface, just model it flat, subdivide it, then bend it.
danpowell.artstation.com
Hammerhead VR - Junior Artist
Basically the shape is just a extruded diamond where you take half of some of the cross sections and move them a bit to the side. It's very simple and I strongly recommend you figure it out yourself. Keep in mind this is much more about being able to see the shape than being able to model it, so trying to paint wireframes by hand on the photo might help you.
Wires here
This is the closest I ever managed to get to the anime references. The cross sections were almost there but the round tip is another conflict area for me. Looking at it now I'm not sure why I didn't use a bend modifier on it. Modeling the prop version I showed earlier seems even more confusing to me though, at least if trying to use subdiv modeling..
The topology is really bad though on this mesh. Every time I look back at the project and try to approach this piece I get a block.
You can set the reference photo as background image in the 3ds Max viewport, apply a transparent material to your mesh, and get a perfect match by placing one on top of the other.
Concentrate on simply modeling without subdivision to nail the nature of the shapes and flow. Once you've successfully done that, apply subdivision and get your hard edges down.
At the moment you're trying to do both things at once, and failing at both. That's unnecessarily complex. Like, would you peel potatoes while boiling them? Peel half the potato, stick that half into boiling water while simultaneously peeling the other half. In theory that might sound more efficient, but in practice it would be a nightmare. Email me for more potato allegories.
It's like you're trying to make mashed potatoes by chewing a raw potato and throwing yourself in boiling water
^ some shape id I might do in my process.
^ I would then start "sketching" each shape while planning on stitching them together
^ Sewing these shapes and adding dimension
^ Mirroring the handle and sewing - You can then bevel or crease (or quadchamfer in max) - bare in mind I was not that careful - you'rs should look better if you take the time.
I have quite new to modeling and have been trying to model this cap from this water bottle. Attached is what I am trying to model and my awful attempt. Really running into a wall trying to model this, any advice would be greatly appreciated.
Technical Artist / 3D Generalist @ Digital Arrow
Thanks again
Technical Artist / 3D Generalist @ Digital Arrow
Technical Artist / 3D Generalist @ Digital Arrow
@rogue1
Process nerd. If you think about things a bit you can break down any shape to it's basic bits - makes it easier for me at least.
Proportions might still not be correct, I was not really going for that! Adjust segmentation for your needs.
...
Im showing a sequence, steps after each other...
- On the first one, we can see a 64 sided cylinder - other segmentations would probably work too.
- On the second one we can see a 1/8th of a cylinder - since the pattern in very symmetrical...
- On the third one we can see that 1/8th cylinder with an edge in the middle... that will serve the base of the detail... proportions are incorrect, adjust height and segmentation for your needs!
- on the fourth one we can see that edge got chamfered
- on the 5th one we can see that the edges got adjusted, i welded some of the bottom vertes to their top neightbour, then I adjusted the remained curve
- on the 6th one, I pushed back the inside verts of the curve shape - very top verts and the ones below that - so I got that flat surface
- on the 7th one, we see no change, I'm showing the shape from step 6 - this should make step 6 easier to understand and implement
- on 8, its just symmetry, copy, rotate, symmetry...
- on 9, I started adding support loops.
- 9, add them inside the curved face
- 10, adjust the support loops that you made inside, so they evenly follow the supported edge
I wont show the top part, that should be obvious
You should get this result after applying subdivision:
Technical Artist / 3D Generalist @ Digital Arrow
Thanks so much
Step 1 make a cylinder with 40 sides or more.
Step 2 the rest ... duplicate rotate with pivot at center ... do the rest
Used this thing to help me have quick smooth (sneaky tease lol)
Will upload the model as an obj for you
If it wasn't clear, my advice was to twist the entire cylinder. That way the intersection lines up perfectly with the underlying structure, instead of being completely at odds with it like in the models on this page. This not only completely avoids any of the aforementioned problems, but is significantly faster to construct.
To make it absolutely clear:
Blue/Red = Other models. Messy, requires lots of manual cleanup and dense geo, leads to bad shading.
Green/Yellow = Clean, logical and efficient.
That's not really right compared to the original image though, the fronts aren't parallel to the edges, its more angled.
Here's what I did. Base 80 Circle though.
Doesn't really matter , I have no whacks with shinny blinn on BUT I do consider that it is not "friendly" for people learning SUBD basics.
And for the method , It didn't took me that much time.
Below. Exactly the same mesh but with the design you mentioned. Makes no difference. To make this change required me to move two edges, everything else is automated. An enormous advantage of keeping it simple and clean. No manual cleanup work. Your mesh doesn't have control loops, btw.
If it wasn't clear, my advice was to twist the entire cylinder. That way the intersection lines up perfectly with the underlying structure, instead of being completely at odds with it like in the models on this page. This not only completely avoids any of the aforementioned problems, but is significantly faster to construct.
To make it absolutely clear:
Blue/Red = Other models. Messy, requires lots of manual cleanup and dense geo, leads to bad shading.
Green/Yellow = Clean, logical and efficient.
Will also point out that even If one would assume It's just an example, It's just not accuracy , In this context we have curvature and a hole ... In your just straight cubes , I know you can make it clean , the point is result on my part , and the fact I don't like to stress U (horizontal direction) and V neither (Vertical) , I assume you would accommodate all these new factors with edges running in the U direction . . .Dude...your mesh is just messy. Accept it. No shame on it...Beeing defensive and argumentative about a method that was JUST PROVED that is less efficient\clean is just....what....i cant even...
If i had to edit that mesh i would tell you to redo it in a way that makes edition on my end easy and not a pain that would require me to redo it...
https://www.artstation.com/artist/joaosapiro
http://www.ieatdeadpixels.com/
So fast and messy = time for someone else to clean up.
Sorry, I posted that before I saw your explanation. (forgot to reload the page)
I'm using blender and use a bevel modifier, so I use that as my control loops and clean up afterwards.
Thanks for clearing things up.
If it were me i would do it in fusion 360 or even zbrush and call it done.
Your mesh would never pass approval here. That might make you upset to hear, and you might feel that I'm a jerk or out to get you, unfair, unholy, sacrilegious or stupid, but it's fact, and one you could benefit from hearing. It's only workflow; you're not married to it.
Shape intersection is a core skill of subdivision and should rely on good practices and experience, not "intuition".
s1dK: No. A base principle of subdivision meshes is to alter the design to align with the technology. Also I can't even begin to understand why you'd suggest ZBrush.
Because It's fast.
ok but approval of what , be precise ? Some HP for low po Bakes ? I don't think you're hostile but Are you saying there's "a way" of modeling? sure nice topology exist, but there's certainly no " way of modeling " for me, and I'm not *cough* the first to say that , And I'm not even a fanatic of a particular workflow, I just do things that bakes right (no crappy surfaces from HP), please guys don't slice my throat I'm being honest.
we understand that it "works" with you , but that would never be approved by me if an artist sent me that .
it was pointed out what could be improved , now what you want to do with that info is up to you , but when you try to convince others of a broken workflow it defeats the purpose of this thread.
https://www.artstation.com/artist/joaosapiro
http://www.ieatdeadpixels.com/
You're absolutely right about that, but once you go ZBrush you can't go back. Iteration on a mesh you've put into ZBrush becomes damn near impossible and could potentially cost you a lot of precious time.
@WayWO It's also important people learn how to make clean subd, and your support loops are anything but in this example. While the others would push for approval on their teams, this is the sort of thing I might buy off a website, and it would be a hard sell with that geometry if I knew I would need to change it to suit whatever I've got in mind. I know you didn't come to this thread with this kind of scrutiny in mind, but hey.
Don`t get me rong, i use sub-d every day but for some meshes that has all kind of intersections like chamfers on top of chamfers it`s best to go with CAD software.
@perna i understand you but if your client wants that exact model you can`t alter the design just becase it`s not possible with sub-d, you must find another solutions to make that exact mesh, there is when CAD software gets prio vs sub-d.
I wonder if the parts below are made in CAD? The combination of very clean edges (non-beveled) with very complex shapes and intersecting model parts seem to suggest a non-subd approach? Or maybe just booleans (although I have to admit that even with a boolean approach in mind, I'm somewhat stumped as how to approach some of it)? It's a garage kit model and I'm pretty sure Netfabb was used in the end part of the process (to cut parts and make molds? I don't really know much about Netfabb), based on some other images from the same person.
I don't know, maybe I'm just a lazy person, but I'm legitimately overwhelmed thinking about how to approach this in subd...
I can't seem to quote individual posts so this will have to do..
I just wanted to chime in on the previous topic.
While it's true that you managed to accomplish the shape, the main problem here is not the fact that your mesh is all messy and shit.
The main problem here is that you have this notion that, as long as you get the shape done it doesn't matter how messy it looks.
In some cases this is true, but with all aspects in life, we should strive to be elegant in our solutions and always be learning more and better ways to accomplish tasks in an elegant, clean and fast way.
This applies to everything in life.
There is always a smarter, faster, easier and elegant way to do things.
Programmers will tell you this. There are coders who get shit done with their messy code with no regards to how clean their code looks... but then if you ask someone to maintain their code.. jesus fuck, what a clusterfuck. Do you mean what I'm saying?
Whilst Perna's post didn't show to to make the shape 1:1, it did show the most elegant solution to accomplish that in a clean procedural way. You can pick up that method and apply it to make the shape the OP wanted.
This is what learning is all about. Trying to figure out the best/smart/easy to maintain solution for if someone asks us to change something, you can do it easily.
I will give you an example it stuck with me when I was in high school. My math teacher once was telling a story when he went to the supermarket to buy bottled water.
The cashier was just a kid, and when he was counting the bottled waters to insert the number on the machine, he counted them one by one...
My teacher was livid. He said, this is why you need to learn the basics and think logically so you put the least effort on yourself. In this case, the cashier accomplished his duty.. he managed to learn how many bottles my teacher was buying.. but he counted them one by one. Sure it worked.. but you know.. he could have just counted the vertical and horizontal number of bottles and then multiply them to get the same result, which would have been easier and would have demonstrated he knew basic logic.
edgesize.tumblr.com » digital playground