Home 3D Art Showcase & Critiques

[W.I.P] ComEng V-Set Train.

polycounter lvl 4
Offline / Send Message
rexo12 polycounter lvl 4
Hello! This is a journal thread for my latest project, the V-set electric trains manufactured by ComEng in the 2 decades to 1990. These are lovely old intercity trains that run in New South Wales, connecting Sydney to the Blue Mountains and Newcastle, among other places. I always found them very charming whenever i've had the opportunity to take them. They are unfortunately being phased out very soon.

I am currently still quite early in the modelling stage:


I will be attempting to use a trim-sheet for this project along the main body of the car, in tandem with modular/bespoke assets. This is my early experiment with laying one out - any tips or comments would be greatly appreciated. Some insight into the best approach to texturing something like this would also be useful.


I expect I will also end up with a decal sheet for details like these:

Some of the modules so far:

I have hit a snag with the pantograph: I have been having a lot of trouble deciphering the mechanics of the thing. This is the model so far:
The lower section of the structure has a lot more going on in my references that I cannot figure out - some kind of spring system that actually drives the pantograph up and down as needed.

Here are 2 of my references. This first image has some kind of double-set of pistons connected to a bellows/accordion of some description - however I haven't the faintest idea how or where either end connects.

This image is of an S/C/K set, which has a very similar design - except the pantograph is not enclosed, affording you a much better look at it. This has a different set of pistons, however even then I do not understand what they do or how they interact. One of them even seems to disappear!


I expect I will just have to camp on a railway bridge somewhere to get the reference angles I need. In the meantime, if anyone has some engineering know-how and could take a crack at explaining how these pistons fit together, that would be much appreciated.

Thank you :)

Replies

  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4

    Small update - replacing the blockouts of the top modules with proper meshes. I have added an enclosure to the pantograph, but the lower sections are still unfinished.

    I think Trim sheets have clicked for me! Seems like a very fluid way to work once you wrap your head around it :)
    In order to preserve texel density I have decided to remove the 2 main tilables from my trim sheet - the Corrugation and main base-metal. These will go to 1-2 MatIDs (depending on whether or not I use MatLayering) on the main body, with a trim MatID for... trims.

    As such, here is the new trim sheet, and how it looks applied to the main body geometry. All WIP of course.

    The question now is whether I should try and extend my use of trims to all of my modules, or whether I should stick to handpainting them. I'm now confident I could get decent results from trims - however I do really love to paint. Unifying everything into a single trim would also allow me to overlay directional dirt and the like more uniformly.

    Thank you :)
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hello again.

    I have painted the doors. I ended up using my trim sheet for the exterior frame, that way I only use a single ID for each door rather than 2 (or more...). Mapping the UVs to the beveled corners is a bit fiddly, and I wonder if there's a trick to it.

    At the moment, the frames have no baked normals and are entirely driven by the geometry/trim sheet. I am wondering if it's worth introducing a 2nd UV channel to map baked textures to, which would also be useful for AO and curvature to drive MatLayering. The painting also needs some refinement, particularly as a lot of my metals don't read as metals.

     I did bring stuff into engine to test layout and materials. I'm quite impressed with UE5 so far, I think one of the more understated improvements they made was really with scene setup and art direction. UE4 was always such a pain for me to get initial lighting setups correct, with a dozen different independent moving parts you had to balance perfectly. UE5 is truly just plug and play..

    I am using a simple objectspace gradient with some grunge maps to lay dust over the top of the model. The front section is also masked out with a Planar mapped stencil.
    Some more work needs to be done on where the gradient fades out, I'd really like something a bit less uniform - similar to how you'd perhaps use a slope blur in Designer.

    Thanks :)

  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Current progress:
    Quickly brought the windows in-engine, and also put together the beginnings of a decal sheet for all the lettering and stuff.
    This is an RGBA set of packed masks, each channel corresponding to a colour parameter in-shader. This lets me re-use and re-colour elements such as the logo, which appears multiple times across the train body. I am considering adding the ability to selectively mask off areas in-shader, for smaller decals where scaling and transforming won't cut it. The logo was sourced from the TfNSW open database.

    The bigger thing I started was on modelling the bogies and undercarriage stuff, something I had been putting off for a while, haha.
    This is just nice and fun subsurf modelling, nothing remarkable here. I'm not completely sure about the proportions but I think i've matched it as best I can - I've started and then restarted this 3 times now. I was surprised at how narrow the rail gauge is compared to the width of the carriage, it's almost half the size!

    I expect I will have a fair few IDs here for the detailed parts closer to the camera, with tiled/matlayered stuff in the back.


    I also came across some photos looking down into the pantograph well, So i'll be able to go back to that at some point. It also showed me some errors I had made with the maintenance bins too :/

    I have also been thinking about how I'm going to present this - I want to build a train station around it. As I see it, my choices are either to pull from somewhere in the Blue mountains, or use an inner city station. This would reflect a tranquil, bright-and-blue narrative and tone or a busy and exciting one respectively. More to come on that.. perhaps.

    Thanks!
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hello.

    Bogie High-poly is probably done, posting here for any feedback before I commit to my low-poly.
    Some personal notes:
     - Some sections don't make much mechanical sense. They are missing motors and hydraulics and links and so forth that sit behind the wheels. I could not find any references of these, and they are hidden from sight normally, so they've been omitted.
    - Very little idea how I'm going to texture this. I expect I will end up with a fair few IDs. A lot of the elements can be mirrored/repeated. This does pose some problems for AO bakes and so forth but It's probably minimal.

    In other news, I have baked and textured the first of the top-bins, alongside the front cabin door:



    I have also tweaked the materials on the side-doors a bit.

    I'm sort of only half-satisfied with the materials so far. They look fine, but they also don't match my references the way i'd like. In particular, that fine brown-dusting that they all seem to have, coupled with very bright and high-gloss metal surfaces is something that the Metal/Rough shader doesn't like! I feel as though I am pulling it too far in one direction or another.



    Thank you :)


  • sacboi
    Offline / Send Message
    sacboi polycount lvl 666
    No crits too offer, I think you're pretty much got it sussed as far as I can see.

    I'm south of the border in VIC and do recall traveling on these trains when posted to Holsworthy Barracks, Liverpool during the 80s, serving with the ADF which if memory serves was kind of weird first time out since we've only single decked rolling stock in Melbourne.

    Anyway will follow along, I like your descriptive commentary in terms of workflow so aside from being quite informative, your project is also a gem for the self confessed railway fan, too :+1:

    Cheers.     
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hey, thanks for the reply. It's great that you enjoy my posts, I'm always trying to provide insights into my own work, alongside any feedback I may receive. I think it's a good exercise for communicating with potential future team members, and probably also helps with actualising a project for yourself.

    It's funny you say that, as as a child I was under the impression that all trains were double-decked. It's only very recently that Sydney has (re)introduced single-deck trains for it's new Metro system.

    Thanks for the encouraging words!
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hello,

    I've been spending time fleshing out the front facade/undercarriage section. Still very much WIP.

    I am basically making much of this up as I go along. I've also been considering putting together a kitbash set of random mechanical stuff to fill in gaps and stuff, I'm not sure how worth it it would be though.

    I did also bake out another of the bins. I expect I will be revisiting these materials.



    Thank you.

  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hello :)

    I have a few things to show this time around - bits and pieces of various parts of the project in their various stages.

    First up is the bogey (bogeys?). I have got a low poly out.
    Blender reports this as 35k tris, although I think the true number is about 64k, as blender doesn't report instanced geometry. UE5 claims it's 40k but I do not believe it.  Regardless, this is much higher than I would like - I was aiming for 20-30k, however I was far more aggressive than usual with optimising this and have already made some significant silhouette compromises. Its going to be a lot harder to rely on trims and tileables for this, as I'm going to need high-res normal maps to bring the high-poly details back, making any potential draw call/density savings kind of moot.
    I am still in the process of unwrapping this, so none of the pieces are consolidated into texturesets as of yet. I try to get each individual piece to an "ideal world" texel density, >20.48px/cm in this case, reducing the density for hidden/unimportant areas per-piece. Once that's done I pack each tile into larger consolidated texturesets with UVP, which is where I sacrifice and normalise much of the rest of the density. I expect most pieces to be at least 10.24px/cm.


    I have been continuing to work on the front facade. Mostly just greebling in a way I hope is convincing.
    Notable changes are: New front lights, with (simple) internal modelling (bulb, reflector). Windshield wipers.

    Pipes! I still haven't attacked that coupling thingo yet.


    I have got around to painting out the trim-sheet,  along with making some slight changes to the base-material.



    My process was to paint onto the high-poly in Substance Painter, and bake all the associated textures down onto a plane in Designer. I had to recombine the maps afterwards, as Designer's baker doesn't really like going from several projected textures to just one. Making changes is easy though as anything I do just gets pumped up and down the pipeline by Substance's ecosystem.



    Changes in-action. I also made some changes to the dust-mask generator. I figured that since objects are separated anyway, I wasn't saving any draw calls by not instancing my trim material, so I introduced some parameters to tweak per-object and better fit the mask to each item. I'm still trying to figure out how to introduce some sharper crevice dirt. I experimented with a 2nd UV map, which had poor results. I also fixed a bug where an unclamped subtraction in the mask was influencing areas it shouldn't have been.

    Have put a tiny bit of thought into the broader environment, with some very rough blockouts.
    This would be based off many of the stations along the blue mountains line, where the way was cut deep into the rock. These sunken stations often felt like a separate 'world' to the surrounding town, as you can't see or hear anything of it once below. There's some interesting opportunities for shadows/lighting here, and it's also convenient for sealing everything in haha.

    At the moment the overpass is supposed to be steel truss and iron, in line with most of the stations. However I have also been considering stone arches, which is probably a bit more romantic. I don't know.


    That is all - This was much longer than I thought it would be! Thank you for reading, feel free to drop any thoughts about the blockout/wider env concept. (Or anything at all!)

    Thanks.
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hello.

    I think i'm satisfied with the front undercarriage bit, I've been working on breaking it down into components for baking and painting. I think all the structural/support stuff and cables will be unbaked, with a trimsheet (costs around 14k), while everything else will be made from a kitbash of painted components.
    This is the full undercarriage module. The big pylon is mainly just to block vision straight through to the other side of the carriage, giving the impression of more density than there really is.


    These are all the components that will be baked + painted. I expect maybe 4-6 texturesets, including trimsheet, for this module. Arguably could be lower but I like to aim for a 20px/cm density to give me flexibility in presentation shots, and this is a portfolio piece.


    You can see the beginnings of the trims/tilings here.



    This is test/unbaked geo in-engine just to see how it works with lighting/if its convincingly detailed. You can also see I have begun painting the bogeys too. It does feel like it's all starting to come together!

    That said, university semester has just started, so I expect my free time to work on this will drop to absolute 0, as my study consumed all my free time last semester, and this time around i'm working!






    Thank you, see you space cowboys...


  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    Hi, small update today. Looking to get feedback on a quick audit of work so far in terms of optimisation & design decisions i've made.

    I have got the low poly ready for the front undercarriage section, and baked/mapped it out. Looks no different to the previous screenshots as yet, so I don't see much point in posting.


    That said, here you can see a bit of a breakdown of the texturesets as they stand on the model, where each unique colour maps to a unique matID (default blender grey notwithstanding). At the moment, there are about 22 texturesets across the entire model, including my trims & tileables, which account for 5 of them.
    My reaction is that this is too much, as i'm really used to targeting like 2-3 IDs for a single 'project'. However, this is the largest single 'model' i've attempted thus far. This has been a natural result of the texel density i'm targeting, and the tradeoffs i've made to balance against polygon counts. I feel as though i've used my trims and tileables effectively, and i am maximising UV space in everything I do, there are just a lot of elements that either don't map to trims very well, need aggressive polycount reductions, or both.

    In terms of polycount it's a bit less clear. UE5 reports between 60-70k at any one time at fullscreen 1440p. I don't believe it considering there's no Nanite elements here and my estimate based on individual props says it should be about 120-200k. 

    Should i be re-adjusting my expectations? Does this seem reasonable for the scale, complexity and target of the model?

    Some insight would be helpful, thank you :)

  • Kanni3d
    Offline / Send Message
    Kanni3d sublime tool


    My reaction is that this is too much, as i'm really used to targeting like 2-3 IDs for a single 'project'. However, this is the largest single 'model' i've attempted thus far. This has been a natural result of the texel density i'm targeting, and the tradeoffs i've made to balance against polygon counts. I feel as though i've used my trims and tileables effectively, and i am maximising UV space in everything I do, there are just a lot of elements that either don't map to trims very well, need aggressive polycount reductions, or both.

    In terms of polycount it's a bit less clear. UE5 reports between 60-70k at any one time at fullscreen 1440p. I don't believe it considering there's no Nanite elements here and my estimate based on individual props says it should be about 120-200k. 

    Should i be re-adjusting my expectations? Does this seem reasonable for the scale, complexity and target of the model?

    Some insight would be helpful, thank you :)


    This is a huge asset - and you're already being quite optimal with it by not using unique textures/bakes, decals, trim sheets, and some shader work.

    If you are considering re-adjusting your expectations, ask yourself this: What this would be used for realistically in a production environment?

    Is it something that's in the far distant vistas? If so, then yeah it can probably be done away with far less geo, details, and intricacies along with everything mapped/baked down into single material.

    OR is it something interactive, player can get real close to it, drive it, onboard it, maybe a focal point in the game where the player constantly has to onboard these railways or see them often? Tons of factors and questions, but if so, a few IDs is definitely warranted for that sort of use, especially that amount of geometry.

    The factors this asset meets (size/scale, importance, screen space etc) all make it quite justified, I think :)
  • rexo12
    Offline / Send Message
    rexo12 polycounter lvl 4
    @Kanni3d That is about what I was hoping to hear, thanks :)

    I started learning realtime art back around the beginning of the 7th gen, so a lot of the benchmarks and metrics I use to measure optimisation is still sort of stuck in that era. As such, I get nervous once I start breaking 50k and 3-4 texturesets (couple this with the fact that i've never tackled anything of this size). So by 'readjusting my expectations' I was looking to see if I should be less conservative, or whether what I had right now was working.

    I understand this is always a bit of a roundabout conversation as there are a lot of factors that go into these decisions that ultimately makes it up to the discretion of the artist, but it's good to know that someone else can affirm my choices.

    Thank you.
Sign In or Register to comment.