Help with molded plastic shapes/transitions (weapon modeling)

Boosted24v

So this was my latest attempt at my Sig Sauer M17/P320 model, which mind you is 1000% better than the last, but, I'm still struggling with the plastic body. I Think I nailed the metal portion, though maybe the topology could have maybe been cleaner. I really struggle with the plastic portion though with its combination of more rigid structure up front with the rail system and the trigger guard cut outs, but the rest is pretty smooth, gradual transitions. I'm wondering how others would approach this, currently I've been blocking it out as one piece and using booleans to carve out the details. Would it be better to cut it up into pieces, like grip, upper square portion and trigger guard and then some how combine them?

My attempt

vs reference

Tiggis

There are more then one way to deal with these kind of shapes, you just have to figure out which one works best for you. Of course you can work on different parts and later connect them all to one single single mesh. If I was doing the frame, I would work on it as a single mesh, focus on the blockout/proportion/larger details/cuts and work my way down to the smaller details. It's easier to work with less geometry and add more when you need it. Having enough reference images so you can see how it looks in all of those angles are important, but also paying attention to them aswell. 

sacboi

Proboolean - Dynamesh workflow https://polycount.com/discussion/168610/3ds-max-zbrush-proboolean-dynamesh-hardsurface-workflow-tutorial

Insights in there that will prove useful, particularly when generating complex objects regardless of technique or app/s utilised.

Boosted24v

I thought I had been doing that work flow, but I guess i messed it up a bit since I was applying every boolean on the go rather than keeping them live until getting into zbrush, that could be a critical change for a better result.

mastering using more geo would absolutely, I tend to create a huge mess when I try to use more, but the lack of geo Is def a key factor in the low poly look I'm trying to get away from.

 

FrankPolygon

@Boosted24v You have a good start but like Mattias mentioned: it's less of an issue with the process itself and more of an issue with the overall shape. The edge width is also a little too tight and a little too consistent across the entire model. This is causing the injection molded parts to read less like polymer and more like metal.

Here's an example of how shape inaccuracies and hard shape transitions can change the way a model reads: Right now the model of the frame has a lot of 90° corners and some of the critical shape transitions are missing. Most hard polymer products are injection molded as separate parts and ultrasonically welded together. Just about every type of casting or molding process requires soft shape transitions and some degree of pattern draft so the part, pattern and mold media will separate without destroying each other. Investment (lost wax) casting can be used to make molded shapes with steeper sides but there's still going to be issues with getting the casting material to flow into the mold before it cools too much.

(Ian and Carl recently published a video about their What Would Stoner Do project and they go over some of the design challenges that go into building up the tooling for the production of two part lowers. Definitely worth a watch if you're interested in how these are designed.)

Sharp 90° corners on plastic parts are relatively uncommon because they can create stress risers and they significantly increase the complexity and cost of the molding process. Adding the pattern draft and minor shape transitions will help increase the overall shape accuracy by more closely representing the actual object. Also note the seam at the parting line where the two halves come together.

When working off reference images it's generally a good idea to start by blocking out parts with known dimensions: barrel OD, bore ID, chamber, rails, etc. then build up the rest of the shapes around the stuff with known dimensions. Looking at some reference images, the guard is about as wide as the top portion of an M1913 rail and the chamfered portion of the rail transitions back into the top of the frame and down into the guard. Adding these minor shape details and shape transitions has a major effect on the overall presentation of the model.

Slightly exaggerating the edge softness of the high poly model is still recommended for a couple of different reasons. As a stylistic choice, it will help catch the light which makes it easier for the player to read the edges and differentiate the surface planes of the object. As a technical aid, it will help keep the edges clean and readable at different view distances and at different texture resolutions.

How soft the shape transitions and edges need to be will depend on several factors:

• Player perspective and view distance.
• Model poly count and texture size.
• Type of material and manufacturing process used.

Here's an example of how edge width and edge softness can effect the way the implied material composition of a shape reads: Tight edges and sharp transitions tend to read like machined metal and wider edges and softer transitions tend to read like molded plastics. Matching the edge sharpness of each material used in the original object and adding some additional softness, while maintaining the unique read of each material, is generally a good strategy. Test bakes can be used to determine how crisp the edges can be before they start to fade away.

From top to bottom: Cage mesh with narrow bevel and subdivision modifiers. Base mesh run through ZBrush dynamesh, group and mask by UVs, 25x polish. Base mesh run through ZBrush dynamesh, group and mask by UVs, 50x polish.

To recap:

1. Study the references and block out the model by starting with items that have known dimensions.
2. Model the detailed shape transitions and add pattern draft to areas that have tapered sides.
3. Replicate the varied edge hardness of each physical material and soften it just enough to support a clean bake.

With practice and time it becomes easier to read and generate the shapes in reference images. It's also worth noting that the Boolean re-meshing process will vary slightly between workflows and programs. Some artists prefer a more non-destructive workflow while others are fine with working in a more non-reversible manner.

If you're still using Blender then it's probably worth running a few tests using the polishing workflow in Victor Kam's Blender to ZBrush guide: https://www.artstation.com/artwork/GXWBoQ This will help you figure out how much you need to polish the edges on the frame to get that injection molded plastic look.

Boosted24v

@FrankPolygon I very much appreciate the in-depth reply with examples, definitely helps with understanding. After reading this I think my technique for this piece was pretty flawed. Starting with block out, id trace a imported picture to create a custom shape, and carve into it with booleans from there. Probably why yours looks super clean and mine is a mess where I'm fighting everything to work. Now I have so many more questions haha! I really apologize for my inexperience I'm definitely trying to learn the right way to do things. As for my zbrushed model I think my edges were pretty polished, not quiet as much as yours but my low poly was way too sharp to catch it, plus the obvious differences between what I had and the actual design which you seem to of captured quiet well.

FrankPolygon

@Boosted24v Not a problem, you're welcome. You're definitely on the right track. it's just about finding a process that works well with your available tool set and getting those fine details down.

There's many different ways to approach creating the base mesh. Some artists prefer non-destructive workflows and others are fine with largely non-reversible workflows. A lot also depends on the available tool-set and what the artist is trying to achieve. As an example: if shape primitives are non-procedural mesh components or if the Boolean solver is struggling with multiple complex shape intersections then it's just a case of having to save an iteration, apply the Boolean operation and move on.

Here's two examples of mostly destructive and mostly non-destructive modeling processes used to generate two different base meshes.

This started out with the known dimensions for M1913 rail and building out a single segment and copying it. The rest is a series of extrusion and shear operations to build up the front of the frame. The circles were used to visualize the segment matching so all of the intersecting geometry lands in the same place. These circles were also used to generate cylinders for Boolean operations. More bevel and shear operations to add shape transitions and pattern draft near the top and front. The remaining circles were used to generate a flat profile of the guard, extruded and a Boolean Union joined this shape into the rest of the frame. The edge loop around the center line was enlarged using a shrink / flatten. Limited dissolve was used periodically to keep the mesh clean.

Here's another example that uses a lot of live Boolean shapes that are controlled by other modifiers like solidify, bevel, etc. Any of the features here can be easily adjusted and all of the destructive editing occurs on the mesh elements that are subtracted via the Boolean operations.

What makes the most sense depends on what kind of software limitations there are, what workflow enhancing add-ons are available, how much reference material is available, etc. The important thing is to accurately represent the shapes and shape transitions while using a reasonable amount of geometry to support the creation of the high poly mesh. Elements of both processes can be combined and utilized where they make sense from an accuracy and efficiency standpoint.

E.g. the base mesh in the first example is derived from non-reversible tool operations but the support loops are completely procedural and infinitely adjustable because they were added with a bevel modifier. This makes sense because the basic shape is mostly flat or angled surfaces and the end goal was a base mesh that's compatible with traditional subdivision. Contrast this with the receiver in the second example which has a number of complex intersecting curves and will be run through a re-meshing process to save time.

It's also worth mentioning that using non-destructive workflow elements can also be very useful if there's limited dimensional information or references and there's an expectation that the model will require a lot of adjustment before the shape is finalized. Again it all comes down to the available tools and the desired results.

Boosted24v

@FrankPolygon I'm still very new to modeling, I got blender because it was free but also got zbrush after gaining a little experience. To be honest I haven't used any dimensions thus far and completely winged it just having reference imagines in the viewport mainly because that's the way I saw it done by a lot of videos on youtube. But would definitely prefer to go the accurate route, especially since accuracy would just make things look better/more realistic and probably come together a lot nicer as well. The dream is to get good enough to make fairly accurate assets for an FPS game based on real world items/weapons. So I'm down for learning and putting in the work to make that happen. My biggest problem probably is that I have tools or methods available to me that I don't even know about yet. For instance I've never seen anyone use a circle as a guide to hand make their own curved edge, I think its pretty awesome, i'd be over here thinking about making an extrusion and then using a cylinder to boolean a curved edge, just because that's all I know. Ill see if can replicate what you've done and see if I cant make a new body that looks more accurate and will do me justice as a base mesh, and again thanks for taking the time to explain different methods an steps.

FrankPolygon

@Boosted24v No shame in using Blender at all. It has it's own issues but it's still a decent piece of software. Especially considering the price.

Overall dimensional accuracy is something that's going to vary from project to project. Everything doesn't have to be perfect it just has to be close enough or reasonable enough. Reference photos in the view-port are definitely useful but there are times when they aren't completely accurate because of lens distortion. That's where judgement and experience helps sort out dimensions and shapes.

One example of this is: most things are designed to whole numbers, rounded decimals or common fractional measurements. Most designers and engineers aren't going to use some oddball dimension like 4.739cm or 29.135°. In both those cases it's very likely the actual design dimension is closer to a whole number. E.g. 4.75cm or 30°. The image just becomes a guide and the model can stray slightly to hit the critical dimensions.

I used to eyeball dimensions but ever since switching to incremental snapping in metric units I don't think I could ever go back. It's just way faster to look up critical dimensions, drop in some placeholder geometry, match that up to a profile image and build the block out around it using incremental unit snapping. This works with both Boolean modifiers and traditional poly modeling.

Just to clarify: it's best practice to avoid generating complex geometry by hand. Manually generating geometry by moving points around is probably the quickest way to get into a world of hurt with hard surface modeling. I'm only using those circles for estimating sizes, visualizing the segment counts and as placeholders for the rotational centers of precision extrusion operations like Blender's spin tool. The circles were also filled and extruded to be run as quick trim tools using Boolean operations in mesh edit.

Here's some documentation about Blender's spin tool: https://docs.blender.org/manual/en/latest/modeling/meshes/tools/spin.html

There's a lot of different tools, strategies and order of operations that can be used to model the same shape. In this way there's not really a single correct way to do things so it comes down to evaluating each process, tool or technique based on its efficiency (time / geometry density) and overall accuracy. The more you practice and the more you research modeling methods the more this stuff will make sense.

Sometimes it helps to use a small segment of a model (maybe something like a section of M1913 rail) and try using different tools and order of operations to generate the geometry and then compare the results. This can often be informative without having the high stakes time investment of completely redoing something before deciding which approach may work more efficiently.

sacboi

Boosted24v said:

My biggest problem probably is that I have tools or methods available to me that I don't even know about yet. For instance I've never seen anyone use a circle as a guide to hand make their own curved edge, I think its pretty awesome, i'd be over here thinking about making an extrusion and then using a cylinder to boolean a curved edge, just because that's all I know.

Can certainly relate too that when likewise starting out but after a while through a ton of practice alongside trial and error, had eventually settled upon a personal workflow that 'clicked' although largely biased toward manual editing since then. However conversely had recently begun to explore alternative non destructive techniques so in addition to Frank's typically insightful solution packed advice plus when time permits of course, I'd also suggest checking out Chipp Walters, NITROX3D tutorial project implementing Blender's modifier suite for rapid visualisation over at BA. 

            

https://blenderartists.org/t/nitrox3d-a-new-hard-surface-workflow-for-designers/1162363

Boosted24v

Much thanks to those who've offered tips and advice, that an combined with hunting down a few videos of people modeling guns on youtube, I've come up with, I think, a better result. Obviously still needs some work but got the major portions of it done. I approached the modeling in a completely different way this time around and starting to realize that almost anything that offers an "easy route" can be a trap. Before like I mentioned I traced out the shape using a blender addon with creates a custom ngon shape which I carved into like a stone with boolean operations. I had watched plenty of videos of people who just started with a plane and made complex shapes just by extruding edges and fitting it all together like a puzzle which is sorta what I did this time. Never tried this before but here is the result. Let me know if there's anything I may have fudged.

FrankPolygon

@Boosted24v A definite improvement to the shape accuracy with the additional details in the front and the softer shape transitions in some areas. Working through the major shapes to make them accurate at this stage should make things much easier when it comes time to add the details later on.

One thing that can be helpful is to analyze the object and break down the surface shapes into two categories: flat areas and transitional areas. This should help identify major shape landmarks and inform how complex shape intersections interact with each other. In the image below the multi colored overlay on top indicates flat areas and the red overlay on the bottom indicates transitional areas.

Every modeling workflow has it's own set of strengths and weakness so each one you learn becomes another tool in the toolbox. The Boolean re-meshing workflow is fast but relies heavily on the accuracy of the shapes used to generate the base mesh. Traditional subdivision modeling (using either support loops or creases) can be a bit slower but it also allows you to control the shapes with a lot less geometry, which can make it easier to change things. There's also cases where the immediate feedback of subdivision modeling can help inform how the shapes will read when the mesh is subdivided and the edges are softened.

Before offering any specific critiques: what areas (with the model or the workflow) have you identified as problems, what strategies do you think would work to resolve these problems and what would you like to see improved?

Boosted24v

@FrankPolygon I've got most of it sorted out I believe, However I did miss that tiny transitional part on the bottom of the rail you marked with purple and have been looking at how to best fix that, for the most part the rest of the rail is spot on. The support has been amazing, I honestly didn't really notice a lot of the shapes until you started breaking things down and has really changed my perspective both on the modeling process and in getting good references. Can't just snag the first side shot an call it a day lol.

I've also been working a bit to clean up the transition where the trigger guard meets the pistol grip, its coming a long nicely.

I'm also super excited about the ergonomic cut out, I was pretty worried about it when I started out, but it came out great. Not only did I not use a boolean to cut it out, which I always did previously. but, its probably the most accurate shape wise that its ever been.

Typically the hand guard on the back was another problem area for me but I think doing the model this way the topology is a lot cleaner and flows properly. Not just me throwing in bevels and fighting it. Sub-d has also been a game changer as well, I used it for the first time after "completing" this model, so rebuilding the body was the first time using sub-d start to finish and the result has be quite a lot better. Other than some minor bumps in the sub-d at some of the transition points I'm pretty thrilled with the difference thus far.