[FINISHED] - Vehicle - M551 Sheridan Tank

Temppe

*EDIT*

Finished model:

Original post starts below.

*END EDIT*

Hello.

I'm working on a model of an M551 Sheridan light tank at the moment. Right now I'm doing the high poly, and have nearly finished the hull. Then I have to do the turret. However, I want to change the weapon system so once I have everything else finished (in terms of the high poly) I'll design a sort of sci-fi energy weapon that looks like it was retro-fitted onto this old tank. Anyway, I'll cross that bridge when I get to it.

This project is to flesh out my portfolio. The reason I want to add the sci-fi energy weapon is because I currently have no sci-fi/makey-uppy stuff on my portfolio. You can see my other work on my Artstation portfolio. Right now I can work on this for about six hours per day so progress should be rapid-ish.

In my usual style I'm using a combination of poly modeling and subdivision within 3ds max and occasionally diving into Solidworks for some components. Obviously, the turret ring in the renders is just a placeholder.

CybranM

Looks good so far, especially the turret ring :>
I'm looking forward to seeing where it ends up

Temppe

Cheers CybranM.

Today I finished the top of the hull.

And... made a start on the turret. The turret is only blocked out, but I did get the mantlet done. I don't have many photographs of this part so it was slow going. Looking back I should have considered making it as a solid in Solidworks but I used subD instead... so I had to wrestle topology as well as the lack of reference. Anyway, I think I can call it done. Well, maybe like 97% done.

By the way, all these high poly parts will get a nice ZBrush pass before baking.

Temppe

Well, it turns out the mantlet was nowhere near 97% done. I did a lot more work on it and the box that sits on top. The side of the mantlet with the gun sight (I assume that's what it is) is a massive headache because of the lack of reference. It's frustrating. I'll have to accept for now that it's not going to be accurate.

Zi0

Looking good so far

Temppe

Yo!

I have been working hard at this project but had a lil' problem with my login in here at Polycount, so I couldn't keep the thread up to date.

I've more or less finished the high poly in Max.

As I said at the start of this thread I made up a new sci-fi type weapon system. What was the IR light (I assume) on the original has been replaced with a target aquisition unit. It was inspired by the radar for the Rapier AA missile as well as the nose of the AH-Z1 attack helicopter.

The weapon itself took inspiration from CRT monitors and also the phased array radars found in the nose cones of fighter jets. I took a few other cues from those telescopes they use to track rockets during launch. As you can see I didn't go overboard with detail since most of this won't be highly visible.

There are still some bits'n'bobs to do that will be attached to the turret like fuel cans, shovels, and things like that. I'm not sure if I'm going to put the commander's machine gun in front of his hatch. It really depends if I can find decent reference. The examples on Prime Portal are all different to each other, and it wouldn't suit the narrative that I have in mind to have this style, in any case. But the heart of the problem, really, is finding enough photos of any one style of machine gun mount.

For now, I'm taking the turret into ZBrush to do all the welds and such. I had a few hiccups getting it in but I've figured out what the problem was. I'm using the new improved chamfer modifier in my workflow a lot now, instead of turbosmooth. But, that generates plenty of NGons, and these were causing all kinds of headaches when they came into ZBrush. The simple solution is to just throw a turbosmooth on after the chamfer and it turns the NGons into quads. Problem solved.

Temppe

Few welds beads...

CybranM

looks good, I like all the small attachment points

Temppe

Cheers.

The welds on the turret are complete.

Temppe

And the hull welds are done as well.

On to the wheels and tracks. I think I'll do around three wheels and maybe ten tracks and repeat the UVs to bump up the resolution. The exact number depends on what fits nicely into a square UV space.

Temppe

And just one more render before bed...

Temppe

Here is a test bake of the treads.

The on e on the left is what I modeled and unwrapped the first time. It had 332 triangles. I optimised it and created the one on the right: 290 triangles. But there are 180 of these in total, so that's 52,200 triangles in total... just for the tracks. I came across a thread where the triangle count of 93,000 is quoted for a tank in Hell Let Loose, which came out in 2018. With this in mind, 52,000 for just the tracks is not acceptable. I'm thinking that 15,000 or 20,000 is more like it.

https://polycount.com/discussion/141061/polycounts-in-next-gen-games-thread

I do seem to remember coming across the figure of 140,000 for a tank in some other game... but I can't remember what it was. Does anyone have any thoughts on the appropriate triangle count for the whole vehicle, and the tracks? Keep in mind that I don't have a particular game or platform in mind, but I want give myself a realistic budget at the same time.

Temppe

And just now I came across an tank asset from that game here and the figure is 40,000 for the entire thing!

https://www.artstation.com/artwork/0yEe4

rollin

Well, make a next gen tank.. I guess if you properly optimize it there is nothing wrong with just pushing the limit a bit and aiming with a higher polycount.

Looks sweet so far btw!

Temppe

Well, I managed to get it down to 120 (120 x 180 = 21,600). So I think that's good enough.

CybranM

War Thunder (and probably World of Tanks) use stacked and tiled poly strips for their treads but for doing them individually 21k doesnt seem unreasonable.

Temppe

Thanks for that resource. I actually kind of wish I'd come across it before I started. Although I did say I didn't have any particular type of game in mind, I suppose I was thinking more of a FPS type game rather than an open world* type game. Which means that I can justify a slightly higher poly count, I suppose. Still, the wiki would have been a handy guide.

*Don't know if "open world" is strictly correct but you know what I mean. Anyway I did more test bakes on low poly parts today.

Bakin' (With the Miles Davis Quintet).
On the left is the high poly, middle and right is low poly. I have added a couple of loops to the corner of the upper component since this screenshot. And I've projected the washers onto the parent part rather than the bolts, if that makes sense. So they'll appear round rather than hexagonal.

Kanni3d

I wouldn't get super caught up in polycount to the point where you're sacrificing fidelity/quality of your art to hit some unknown 'budget'... Yes, be reasonable with how you spend your triangles, but as someone mentioned earlier - if you're pushing quality, then its justified to have a high tricount.  Great work so far

Temppe

Hey hey guys, thanks for the encouragement. I'm still chugging away at this.

I'm unwrapping and test baking as I go. And fixing all the little mistakes I find at the same time. It's getting pretty tedious now, with the Max crashing frequently and lots of hopping between heavy files... gotta grin and bear it.

Or grin and beer it... lots of beer.

Nah, I'll save the partying for after.

Temppe

Back with another one of those block rockin' screenshots. Two of them, in fact.

These are the panels that cover the engine at the back of the hull. I think I'll apply an alpha map to the wire mesh and model some very simple fans underneath. I think basic geometry should suffice, no need for a normal map or anything since they will be party obscured - that way I don't have to give much UV space to 'em.

Here's a close up.

I've come to realise that I made a rather shitty mistake while making the high poly. I used a few bend and noise modifiers and slightly randomised the transforms of some subcomponents of parts to add to the realism. As I kept modeling I ended up collapsing these subcomponents together into one part - thus deleting the modifier history and baking in the deformations. This had made it a pain in the bollox to align the low poly with the high poly. And in some cases it offsets my sculpted weld seams. I had to correct the weld seams in some cases with FFDs, which works but of course it's not going to be perfectly blended into the parent surfaces like it was in ZBrush.

Reaslistically I'll get away with it, considering how much resolution these parts will get in the final model. If I see that it looks shit I can FFD it some more. Anyway it was a massive time waster and a lesson well learned.

Temppe

Yo, quick update though I don't have any new images (I might later tonight).

Today I had a bit of a meltdown because I noticed that I had to redo a lot of work. I'm not sure what the reason is, but it must be either file corruption during a crash (happens a LOT with this project), or carelessness by yours truly. Basically, some objects  have gone missing into the aether, some have become attached to others, some have lost sub-objects... The autoback files were too recent and my own backups were too old. So I spent the morning wondering if I should cry or punch a hole in my monitor. Then I spent the afternoon sorting it out.

Anyway, here is what I learned:

1) Use layers. Always. This prevents accidentally attaching the wrong object to another and not noticing.

2) I need to update my bits'n'bobs library. Actually not update it, but completely overhaul it.

3) I need to clear out my autoback folder and have more autoback files on rotation for a project this size.

4) I need to separate the major components into different files while I work on them. Having the program pause for a full minute while it autosaves every ten minutes really sucks. Also just the constant, general lag really compounds all of the above problems.

5) I need to totally reorganise the text file where I record all the lessons I learn on each project because right now it's a total jumble.

Aside from that, I have to start thinking about how I'm going to organise the texture sets. I notice on this impressive piece on Artstation that they used six 4K sets.

https://www.artstation.com/artwork/aR6Omq

That's probably excessive for my project. I'm thinking that separating all the little protrusions on the hull and turret into their own set might be a good idea. And then have on for the running gear, one for the hull and one for the turret. And the set for the protrusions wouldn't have to be 4K.

Any thoughts?

CybranM

I feel for you, losing progress always sucks.

Your fourth point about separating the project into different saves, its probably a good idea although it gets a bit more annoying to combine them to get a good overview. The time saved by less lag and shorter saves is probably worth it though.

The artstation link is broken so I cant comment on that.
With the small protrusions and such, I'm no professional vehicle modeller but there's usually space in between the larger UV objects to squeeze in a bunch of small stuff. My understanding is that if you have transparent elements (engine mesh etc) you'd want to combine all of those object into the same texture to save performance in-engine.

Keep it up, its looking better and better!

Temppe

Thanks CybranM. The link is fixed now. Also good point about separating the parts with the alpha map - I hadn't thought of that.

Here's a preview of the hull. Note that the UVs are not optimised here, I just used the standard Max packing tool. I've played around with Unfold3D in Maya and it usually does a decent job of fitting in the bits'n'bobs between the larger elements - though it has a habit of leaving medium sized areas at the top of the map unused. In my experience I can manually fill these areas up though, by rezising some islands.

Also one more thing - the gaps will be closed after baking.

Temppe

Moving right along, ladies and gentlemen...

Temppe

Time to start painting...

Sage

Love the details on this model.

Temppe

Thanks Sage.

Here is the first pass in SP. Still a lot to do as you can see. I'm not sure about the colour of the radar and weapon. The idea is to indicate that these are experimental and are being tested on this vehicle. I tried making the radar white but it didn't really work... I'll keep playing with it. Also, throwing on some text graphics might get the idea across as well.

I also have a couple of questions to ask but I'll do that tomorrow morning - right now I need bed!

Temppe

Yo, this image is actually from yesterday. But Photobucket is acting the plonker so I can't upload a more recent one right now...

Temppe

Hmm... Dunno if these links are working right now... not impressed with Photobucket.

CybranM

Yeah you probably shouldnt use photobucket. I'd suggest imgur or directly uploading them to polycount.
Its always a shame when you go into older threads and all the photobucket images are gone. I'm not sure if they auto-remove older images from their servers but its damn annoying.

Temppe

Thing is I actually paid 'em eighty dollars last April for two years of hosting so this situation would not happen. Maybe I should get onto their support, because frankly right now they're robbing me blind.

Anyway, I realised a few hours ago that I forgot to model the low poly sprockets. I just finished them but it means I had to redo the UVs for the tracks/wheels. And that means I have to re-bake and re-paint as well. Thankfully it should be a pretty rapid process, the way I have the layers set up in Substance.

Temppe

Okay, Photobucket seems to be cooperating again... for now.

I've pretty much finished this guy. The only thing I want to add is a couple of accessories like the shovel on the back of the hull. I have plenty of space on the UVs with the basket mesh behind the turret.

I'll explain the narrative that I had in mind when I made this project:

Around 1980 the United States and West Germany cooperated on the development of a directed energy weapon for the anti armour role. One of their prototypes was mounted in a Sheridan light tank and trials were conducted in West Germany. Ultimately the Americans withdrew from the project and, without funding from the US, the West Germans halted the trials.

This vehicle sat forgotten for some years in an army base until 1984, when the ultra left militant group Red Army Faction raided the base's armoury to obtain weapons and found it. Obviously they were pretty happy with their find...

Anyway, bit of a silly story but there you have it. I should have this project complete by tonight. Accessories, posing the turret and some lights... finished.

Temppe

I'm calling this baby finished.

More renders will be available on my Artstation:

https://www.artstation.com/artwork/VdXLP5

In the end I decided I'd get more value from the remaining UV space I had on a fuel can rather than the shovel.

Triangle: 164, 237

Texture: 3x 4k (albedo, normal, rough, metal) and 1x 1k (albedo, normal, rough, alpha).

I wish I could say that I enjoyed this project and learned loads, but to be honest I feel like I have just as many questions now as I did at the start. There's a list as long as my arm of things to test. And while I'm very happy with the result I have to say that the last few weeks of this project really pushed my workstation, and dealing with constant crashes and lag was not an enjoyable experience.

I think it'll be a while before I take on a project of this magnitude again. Unless someone's paying me of course!