Game Assets With Fusion 360

yeah this might be a weird one we're going to try making a couple game assets with fusion 360 and some other tools most of this video will be a modeling overview of a couple random models but the purpose of them is a bit different here we're going to try making some low or lowish poly assets that we can bring into a game engine in this case it's in unity for a vr game i'd say this is less of a tutorial and more an overview of some experiments that seem to mostly work anyway let me back up a bit for whatever reason i've been trying to learn how to make vr experiences in unity and this latest attempt started by trying to be like this old arcade game super off-road essentially just trying to be a vr racing game that doesn't give you motion sickness i was curious to see how far i could get if i dropped maybe 60 to 70 hours or something into it but rather than do things a normal way i figured since i'm more comfortable with parametric modeling than poly modeling why not try making these models in fusion it's not really made for this type of thing but i figured it should work and i'd say that looking back for the most part it does so this was the end result it feels almost like a real rc car and you can race against yourself from the past it's better than nothing i guess and most importantly it has a weird random spike hand that i could poke things with and i suppose it works for pressing buttons too so here's the idea what we're going to try for each of these is to make two models one high detail one low detail then bake the details from the high onto the low and i'll be doing this in 3ds max but you can do the same in blender or just about any other 3d animation software it's pretty standard practice so there are tons of tutorials on this topic it's just that the modeling tool we'll be using is a bit different than usual let's start with the controller since that's the one i tried first i started by blocking out a sketch profile that would drive the rest of the geometry there wasn't any design in mind here just looked at a few photos and kind of made it up as i went since this was meant for vr i tried to model this to roughly accurate scale when modeling i'm constantly referencing either a set of calipers or a tape measure to make sure i'm on the right track then i started blocking out some of the main shapes by extruding the profiles to give a little more realism there's a slight draft angle from the center parting line once these blocks were in place i jumped into the sculpt environment to rough out some kind of a handle in reality this would be asymmetric but decided to use symmetry to save some time i was a bit lazy on this one so i didn't plan it out as well as i could have here the main idea is to get the driving face loops in place then fill in the gaps if you watch the time lapse for this which i'll also upload you'll notice that in the place where i didn't do that somewhere around here is where i stumbled a bit and i ended up deleting an entire edge loop to clean up the model a quick tip here is that for refining organic shapes like this it's sometimes helpful to work in view space coordinates that way you can quickly move things around based on the viewing angle another approach would be to use multi-view but for me the view space trick is a little faster and easier the transition from the handle to the rest of the model was done with patch tools created a couple split lines that would help transition to the geometry above it a couple extruded surfaces helped to maintain continuity with the symmetry then a couple lofted surfaces that were trimmed back using the driving sketch for this feature and finally a patch to fill in the middle this isn't the cleanest approach it's not fully curvature continuous to the upper section because the surfaces we started with weren't curvature continuous from the beginning causing this transition to break any attempt at a g2 feature since this is for a game that nobody will play nobody's going to notice but something to keep in mind if you're making something for real then we stitch everything together into a solid and combine it back with the rest of our model the rest of this base model is relatively straightforward essentially just extrusion after extrusion referencing the initial driving sketch and blocking in the other main geometry once this was at a good enough place i made a copy of the model and started combining pieces together ending up with three bodies one for the static elements which is most of the controller then two more for the elements that would move so the trigger and the wheel now we can jump back to the other model and start adding details most of this comes in the form of fillets and tiny extruded details anything that was about one millimeter or less in terms of fillets embosses d bosses splits or anything else would end up making it onto this high detail model and i also use this opportunity to add splits for material breaks that's pretty much it for this model moved the low detail version back to the origin and exported everything as a step file look at exporting in a little more detail later on alright so the car was a lot of the same story start with the base model make a copy of it then add detail to one of them this wasn't any more difficult than the transmitter in fact most of it was easier there's just a lot more of it i like to start with the wheels because it gives an easy reference for a scale and proportion then we can work our way from the wheels inward there was a lot of lazy modeling on this because i didn't want to spend more than a day on it so you'll see things that you shouldn't really do with parametric modeling like making copies of bodies and just eyeballing their positions so we've got some wheels in place and we're going to rough in a couple base sketches similar to what we saw with the transmitter just to keep us on track and help with proportions then we can start blocking in parts from the wheels and working our way to the center again this is just extrusion after extrusion with this one it was particularly helpful to work parametrically because since i was just doing everything by eye i wasn't sure where parts would end up relative to each other this way i could easily go back in the timeline and make changes to have everything lined up correctly you'll notice that i placed a simple extruded cylinder in place of the spring for the shocks this is because the spring would make for a rather heavy mesh so we're instead going to try baking a spring onto the cylindrical profile the body was a hybrid t-splines and overbuild service modeling approach i tried to make it look kind of like the real rc cars that are out there now sort of a frankenstein with cubes from different brands this means a lot of crisp edges which i felt was easiest by making multiple surfaces using t-splines and doing a lot of surface trims i had more time i would have spent a lot more effort on the design here so once the body is done that's it for the base model and we can do the same as we did with the transmitter make a copy of this combine the static elements and simplify the model again going and detailing one of the models adding fillets cuts and all that if i were to do this again there were a couple things i'd change the first is the rib details on the inside of the wheels these details were too big for the normal maps to make sense of i guess i knew that when i was modeling it but i figured they wouldn't be visible enough to worry about this is about half true because i do notice it went up close the next thing is that i spent virtually zero effort in trying to get the color splits on the body looking good it would have been nice to actually design an interesting paint scheme maybe put a few graphics or logos or something now that that's all done export this again as a step file and that's about it for the modeling process so let's talk a little bit about exporting and various formats you might consider there are probably a dozen possible paths we could take with this what we have here is a solid b-rep model but unity unreal and other game engines only run on meshes now there are importers like pixies for unity and datasmith for unreal that can import solid cad formats like step or set files but there's a few things we may want to do to these models before bringing them in so let's say for now that we want to bring this into some kind of 3d animation software like 3d studio max blender maya c4d or a few others so that we can do a bit more setup a couple of these tools like 3ds max or the studio version of cinema 4d can import cad formats and handle them pretty well since i use 3d studio that's why i exported a step file i like doing it this way because it gives me a lot of flexibility even after import but not everyone has access to 3ds max i may not in the future either so let's briefly look at what we might consider if we were exporting to something like blender for this we need to create a mesh file there are a few options for this from fusion we can export an obj an fvx or an stl file unlike stl though obj and fbx files allow you to easily export multiple bodies like we have on this model to one file but obj and fbx files don't give us any export options for things like mesh resolution like we have when we export stls but there's a way around this that just takes a few steps first and this is optional i like to create a copy of this fusion project just to keep the original clean then we can create a mesh feature and use the b-rep to mesh tool to convert our geometry to mesh bodies this will give us the same level of flexibility as we get when exporting an stl file we can create a high and low res version for example one trick you can use if you have a lot of objects is to select all the bodies you want to convert in the model tree before entering mesh mode this will save you from having to individually select bodies we exit this mesh feature and now we can remove all the non-mesh elements and export this as an obj or fbx and bring it into blender or anything else i generally prefer fbx because of how the hierarchy and material groups are brought into blender and just note that you may need to apply an auto smooth to normals in order to get the mesh to look right so we've got a few mesh files now in our 3d animation software for the rest of this i'll be using 3ds max and unity but the workflow will be similar no matter what you plan on using i won't go super in depth here but we'll try to get the general idea so it can give a few topics to search for if you want to do something similar while we're in here we're going to do a few things we want to do a full uvw unwrap of the low poly version because we will need this when we make the projected maps we can apply some textures to the high poly version and this would also be a good time to apply logos and labels or whatever i'm using these maps mostly just to make areas of color onto the model so it's just a few standard scanline materials with self-elimination turned all the way up this way we aren't capturing any of the lighter shadow then we're going to use the render to texture tool with projection mapping enabled to make a few maps on the low poly version and this tool is apparently redone in the latest version but i haven't tried it yet for most of these i'm only capturing the diffuse channel which contains the surface color and the normal which is what allows us to grab the details from the high poly and for the transparency of the spring i just did this manually in photoshop by modifying the diffuse map if we were using something other than scanline for this we could have also made an occlusion map maybe a specular map but i thought this was good enough so with these new maps created we can see how close to low poly versions are to the high poly even with maybe 1 20th the number of faces or something like that so we've got these models with textures but there's one more thing we want to do with the car ideally we want to keep the number of meshes to a minimum in order to limit the number of draw calls especially if we're building for the quest but we've got a full suspension with springs and arms and all kinds of things that need to move so what i decided to do was keep everything as one single mesh body but create multiple morph targets for how the suspension would move we could then apply various weight values to those morph targets based on the relative positions of the physics colliders we'll apply in unity this actually worked better than i was expecting after messing around with the physics parameters for what seemed like forever we now have a car that feels and drives somewhat like a real rc car we've also got a controller and it does stuff so now i guess we just need to make a track and an environment and a ui and make it register lapse and get something to race against and make it generally not terrible we can look a little bit into the environment and the track because they had a slightly different approach from the car and the controller the environment was again modeled in fusion 360 but instead of baking from one object to another this time i used octane to bake the lighting and textures onto a new map channel and the graphic on the back was some lazy node box work it's a fun little free program if you haven't tried it the track was made in mudbox from a simple plane for the most part i just used two brushes the wax tool and the flattened tool then just painted the dirt texture onto the track and exported this went pretty quick so it should be possible to make a whole bunch of these if i get really bored one day there's a lot of things i could have done differently aside from generally having better lighting and graphics and everything probably the biggest thing is that the car has all this detail but it's always so far away it's only ever 20 pixels wide or something like that so it should be possible to greatly reduce the detail without losing much another is optimizing the use of the texture sheets i think the way they do this properly is to shrink all the things that need less detail like the hidden parts of the chassis to allow other things to take up more space on the sheet allowing us to capture the more important parts in a higher resolution there are also a few challenges with smoothing when it comes to dealing with converted cad models most of them aren't deal breakers and it may be better with a different render pipeline but they are noticeable if you look for them but yeah i guess it works i i think i need to go outside for anyone interested i'm also going to upload a time lapse of these models so you can see how they were put together they won't be narrated and it's sped up to a point where it's virtually impossible to tell what's going on but it'll be there if anyone wants to see it until next time with something probably completely different you

Channel: name notimportant

Views: 4,274

Keywords:

Id: w7K9qLeWhKA

Length: 14min 47sec (887 seconds)

Published: Sun Sep 27 2020