Welcome to! Click one of these buttons to join in on the fun.

3DU.S.S. Coronado, Katana Class Starship



  • evil_genius_180evil_genius_1801843 Posts: 9,995Member
    That's some great work. I like how much thought you're putting into launching and landing the Specters.
  • McCMcC346 Posts: 697Member
    edited August 2019 #153
    If you're just joining this thread, hi! Please read the first post for some background info. Many questions have come up over and over again that are answered there. Thanks!

    Thanks, @evil_genius_180 !

    Was out of town for GenCon (my first time!), and the week prior to it was a whirlwind of in-person D&D sessions, so have only really been able to get any progress done in the last week or so.

    The next big item on the agenda was finalizing what to do with the actual launch doors. After some experimenting, I decided I liked horizontal staggered sliding panels. Unfortunately, in order to make them not overly huge and also reasonably thick (~30cm, my standard hull thickness), I had doors that didn't fit into the door frame I had made! Time to rebuild the entire outer shell of the Specter bay, I guess!


    Pictured above is the "how am I going to patch this up?" screenshot of despair on top, with the end result (using a red carpaint matcap to look for mesh distortion) on the bottom. The Shrinkwrap modifier, using some freshly-built and very clean target meshes, made reworking the area a lot less painful than it would have otherwise been. That'd turn out to be foreshadowing for later tasks. Speaking foreshadowing, anyone remember this?
    McC wrote: »
    Shrink Wrap is a dirty, evil modifier that I try to avoid wherever possible. opinion on that topic has changed over the years. :sweat_smile:

    In any event, I ended up putting together a quick rig and cooked up a viewport render to see how it looked opening.

    Post edited by McC on
  • evil_genius_180evil_genius_1801843 Posts: 9,995Member
    Looking great. Those are my favorite kind of bay doors to do.
  • BrandenbergBrandenberg1487 CaliforniaPosts: 1,858Member
    Fun. Click and open the doors. It's looking great.
  • McCMcC346 Posts: 697Member
    If you're just joining this thread, hi! Please read the first post for some background info. Many questions have come up over and over again that are answered there. No, this is not the Insignia class. Thanks!

    Thanks, @evil_genius_180 and @Brandenberg!

    Still at it! Also, I'm a dad now!

    I've been posting periodic updates in Discord since I've gotten back around to working on the model again (after another half-year hiatus!). The new Intel denoising node in more recent versions of Blender has more or less solved my demoralizing noise issues entirely, which is awesome. I set about the arduous process of UVing the main hull (the saucer has been UVed for a while, as evidenced by the multiple color tones on it). I should've kept up with my UVs as I went, but I got lazy, and now I'm paying for it!

    In the process of doing so, I started seeing more and more mesh distortion, especially around the windows in the main hull. I knew if I didn't address it, it'd eat at me forever, so I rebuild every single offending panel with significantly higher polygon density and vastly superior loop flow, then re-integrated it into the existing (and now quite a bit lower resolution, though it doesn't show thanks to the placement of the panels) hull. Finally finished that this evening, so I can now return to UVing and--hopefully soon--actually do some texture work and, maybe, possibly, finish this project at last.

    Some "establishing" renders from Feb. 8th, just to refresh on the overall look of the ship.
    coro_2020-02-08-1138.jpg&size=320 coro_2020-02-08-1140.jpg&size=320

    Added an underside fill light on the 10th, no longer lit only by the single sun-bright key.

    After chatting with @Viper on Discord, switch to the same HDRI he uses in his WIP renders on the 18th.

    Before/after comparison of the hull panel rebuild and the comparative mesh density from yesterday evening.

    Render of the finished panel rebuild from this evening.

    Not-quite-the-same-but-close before/after comparison render from the same angle as above of the overall wires of the ship from last summer and now.
    coro_2020-02-23-2338_wire_old.jpg&size=320 coro_2020-02-23-2331_wire.jpg&size=320
  • BrandenbergBrandenberg1487 CaliforniaPosts: 1,858Member
    A DAD!!! Boy is YOUR life going to change.

    Keep up the good work Dad
  • McCMcC346 Posts: 697Member
    edited March 2020 #158
    If you're just joining this thread, hi! Please read the first post for some background info. Many questions have come up over and over again that are answered there. No, this is not the Insignia class. Thanks!

    Thanks, @Brandenberg! Life has change profoundly, but also remains the same in a lot of ways!

    Been madly UVing since last post, finally getting everything to a point where I think I'm happy with it. With the UVs sorted, I plunged back into texturing in earnest. Right now, I've got a generic hull panel mixed into the shader that is just mapped with blended box mapping. I don't intend to leave it in that state; the hull panels will get properly incorporated into the textures once the other features are in. For now, it's there just to give me something to reflect light off of and see how the surface roughness is coming together.

    UVing, from Feb 26, 27, and 29, using checkerboard textures to check scale across both islands as well as different UDIMs.
    coro_2020-02-26-1514_ogl.jpg&size=320 coro_2020-02-27-1316.jpg&size=320 coro_2020-02-29-0932_ogl.jpg&size=320

    Texturing finally beginning anew (albeit with the previous texture work used as a reference), from Mar 1. Some pieces still have checkerboard scale textures on them, other places have islands with solid, bright color for quick identification.
    coro_2020-03-01-1120.jpg&size=320 coro_2020-03-01-1135.jpg&size=320

    More incremental texture work, from Mar 3 through this evening.
    coro_2020-03-03-1320.jpg&size=320 coro_2020-03-04-1154.jpg&size=320 coro_2020-03-07-1251.jpg&size=320 coro_2020-03-07-2333.jpg&size=320
    Post edited by McC on
  • McCMcC346 Posts: 697Member
    edited March 2020 #159
    If you're just joining this thread, hi! Please read the first post for some background info. Many questions have come up over and over again that are answered there. No, this is not the Insignia class. Thanks!

    Texture, texture, texture! Plus messing around with shaders and not being happy with any of my experiments. I've (at least for now!) resolved to leave the shader fairly simple for now and will circle back to roughness and bump maps after I get all the colors and such sorted out.

    Most of the texture work is done in Inkscape (open source equivalent to Illustrator).

    Mar 10:

    Shader iteration over the course of Mar 10 through Mar 14. I have yet to shake the feeling that it looks too much like a plastic model or toy in the shots, which is super disheartening:
    coro_2020-03-10-2255.jpg coro_2020-03-13-2233.jpg coro_2020-03-13-2306.jpg coro_2020-03-14-1418.jpg

    Phaser hazard markers from Mar 15:

    Blocking in some more color areas and general markings, from this evening:

    I'll probably back off of the big dark landing area aft of the shuttlebay. That was lifted from the aft bay of the Sovereign-class, where it's a much smaller accent. Might keep some of it, though -- maybe make it U- or V-shaped instead.

    The grind goes on.
    Post edited by McC on
  • evil_genius_180evil_genius_1801843 Posts: 9,995Member
    That's looking fantastic.
  • SATRSATR135 Posts: 317Member
    I live the deflector dish
  • P5ych0p4thP5ych0p4th444 GermanyPosts: 336Member
    edited March 2020 #162
    McC wrote: »
    I have yet to shake the feeling that it looks too much like a plastic model or toy in the shots, which is super disheartening

    Well that kind of depends on, what you‘re going for. I for one like a more plastic-y look, since it makes the thing look more like a super detailed physical model, that an actual spaceship. And that in turn tells our brains, that it‘s „real“ ... weird, I know 😁

    For example: a surprising amount of baby Yoda is cgi. But they went for a more plastic-y look, instead of a skin shader, so he looks real to us.

    Almost forgot to write: your work looks amazing! 👌
    Post edited by P5ych0p4th on
  • McCMcC346 Posts: 697Member
    Finally time to post a bit of an update, albeit not the kind folks generally hope for. Between April and August, I didn't have any real time to make progress. Work has been pretty insane this year (and it's only getting crazier as the year goes on), on top of raising a newborn and, uh, a global pandemic. But as far as Coronado is concerned, things were looking even more bleak.

    Around the end of August, my largest desktop hard drive started to fail -- hard. We noticed it when my wife and I sat down to play Divinity Original Sin 2 multiplayer together for the first time, and the game just stopped loading. Killed the process, went back to Steam...which told me the game wasn't installed. When I was literally just loading into it. Come to find out, all sorts of things are suddenly inaccessible. Spent a bunch of time trying to rsync would I could off of the drive, in the hopes I could rescue it before it died. Spent a bunch more time trying to use low-level linux processes to extact more (mostly to no avail). Finally reached out to a coworker who had also had a hard drive fail on her and had had good results with a particular data recovery service and got the ball rolling with them.

    This drive contained, among other things, my 3D art projects (including Coro), my writing, my music collection amassed over the last 20-odd years, and so on. But because I'm a genius :eyeroll: , it also contained my "master" git repositories -- i.e. my backups. I had some of this in other places, too; the writing and the git repos for it live on pretty much every computer I own, so I didn't really lose anything there.

    I'd had this drive for somewhere close to a decade. Therein lie the problem. Until this drive, I had been upgrading computer parts on a regular-enough schedule that running afoul of a catastrophically dead hard drive had just never happened. Finally hit a state of complacency sufficient for time to accrue and a drive to just wear out. Apparently, it took something this big for me to finally, finally get serious about backing up my shit (which I have already started doing with an external SATA drive dock I purchased for just this reason many years ago, but never actually got around to getting into the habit of backing up to -- how's that for a real punch in the gut?).

    The data recovery company managed to (nominally) recover about 74% of the drive. They recovered a lot of stuff cleanly, and more stuff that may be corrupted. Coro's blender files--all of them, every single incremental one--were among the "may be corrupted" group, but at least they were in the list of files recovered at all. Odds were decent, though, that the entirety of the Coro model was gone. The git backup repo for Coro also contained a handful of corrupted objects.

    Tonight, I finally mustered up the courage to dig through all 178 incremental save files that were sitting in the recovered folder. Picking randomly, I received failure-to-load after failure-to-load. Trawling through the files at the byte level, looking for the absence of a particular byte sequence I knew to identify a corrupt file based on the way the data recovery folks did the recovery, I identified 19 files that might load.

    Of those 19, five of them did.

    Coro isn't dead after all!

    It's not all roses, though.

    The most recent of those five files is still about a week and a half "behind" where I left off, and I still have the daunting challenge of getting the git repository back into a functional shape, since so many parts of it are corrupted. I'm tempted to just abandon the history altogether and snapshot what I have, without worrying if any of it is something I could "roll back" to.

    I've also lost quite a few of the master texture documents, since I didn't save those incrementally the way I did the Blender file, but those are also much easier to reconstruct. Many of the SVG files I used as the basis for the actual 4K maps and UDIMs are in good shape, so reconstructing the various layering effects should be straight-forward.

    I don't know how soon I'll be back to WIP-render-ready shape...but the show goes on.
  • ComcoComco299 Posts: 1,272Administrator
    edited October 2020 #164
    Sucks to hear about the lost progress, but glad to hear you managed to get something back, McC. I was enjoying seeing this one come together.

    Please let this be a lesson that others can learn from - it's so worthwhile getting an online backup service. You set it and forget it. But it's saved my bacon a number of times, over the years. :)
    Post edited by Comco on
  • RekkertRekkert2395 Buenos Aires, ArgentinaPosts: 2,119Member
    edited October 2020 #165
    Sorry to hear about your disk failure, I know what it is to lose motivation to work on a project after losing progress. :(

    +1 on what Comco said, I got serious about online backups after it happened to me, using just a simple Google Drive setup saved me a bunch of times already.
    Post edited by Rekkert on
    For my finished Trek bridges and other works of mine, visit my portfolio
    Feel free to contact me if you're interested in commissions!​
  • McCMcC346 Posts: 697Member
    Time to do an actual update now!

    Even before the crash, I made several changes that I only ended up posting on Discord. First, as mentioned in the last post with updates, I backed off on the big dark area behind the upper shuttlebay and instead made it more of a U shape, with additional landing lines and such. This was from March 21:


    Looking at that, it became clearer and clearer that my little "porthole" windows in the insets on the saucer just weren't working, so I replaced them with more Voyager-esque inset windows instead. This was from March 25:


    I also started having serious misgivings about the "spine" and why such a huge dividing feature would run through the middle of an otherwise contiguous flight deck. After a lot of experimenting, I ended up with this more subdued piece. This was April 2:


    Then the crash and such happened. When I finally got back into it, I decided that I still wasn't happy with the above solution and decided to throw the full-length spine out entirely, in favor of having it terminate with the other hull plates in that area. This was October 31:


    That, of course, left a big gaping hole in between the two Specter doors, and someone suggested putting a flight observation room there, so I mocked one up just to see how I felt about it. I wasn't sure straight away, so I let it sit for a bit to think it over. This was also October 31:


    A few days later, we (some of the old Coro RPG folks) got to discussing the flight procedures for the Specters and it turned out many of us had misunderstood how it was intended to work. The "top" doors were meant to be where they launched from and the aft doors were meant to be how they returned to the ship. I had been treating the aft doors as egress, including a little aircraft carrier-like catapult launch track for the Specters. The original intent had been for them to be raised on a variable gravity platform for launch, and then that platform would launch them in pretty much any direction, with huge initial velocity, that the situation called for. This called for a couple of changes to the area: ripping out the catapult "strips" on the deck (which I was all too happy to do!) and restoring the interior elevator platform. I also reduced the overall height of the aft doors, in part because they were playing merry hell with the sense of scale in that area. I ended up going with a revised two-piston platform, drawing inspiration from some of the model kits I've seen of the Constitution refit's shuttlebay and the shuttle lifting platform inside it used to elevate shuttles to the launch deck from storage. This was November 6:


    I revised the shape of the exterior flight control station, which at this point I had settled on as a worthwhile feature to keep. I drew inspiration for the window shape from the flight control station on the Sovereign dorsal shuttlebay, which resulted in the following. This was from November 7 (and please ignore the UV wonk on the spine!):


    And, to sanity check how it looks in deep space:


    At this point, having burned a lot of unexpected time on remodeling, I decided that the model is now geometry locked. I'm not going to make further structural changes, no matter how compelling. The only changes to geo henceforth'll be driven by render errors or UV issues and the like.

    So, back to texturing!

    Most of my texturing source files got corrupted, but most of the textures themselves were still around, so I can reverse-engineer a lot of that without too much trouble. My vector documents for making precise panel lines and things are also still intact. To that end, I decided to just start doing some direct on-model texture painting in Blender to sketch out what I wanted to do with more precision in the final texture comp. The following are the results of the initial work in that direction, all from November 10:

    coro_2020-11-10-2228.jpg&size=320 coro_2020-11-10-2225.jpg&size=320 coro_2020-11-10-2232.jpg&size=320

    The stardrive/engineering hull in particular has only had a little bit done to it so far (top and underside of the pylons and the ventral "notch") -- the lateral/dorsal surfaces remain flat color at the moment, but I don't intend for them to remain so. The big dark patches on the underside of the saucer are meant to be similar to the big sensor pod on the Nebula class.

    Obviously, a whole lot of shader work left to do, as well.

  • psCargilepsCargile330 Posts: 596Member
    Looking good as usual. Shuttle bays look more functional without that spine.
  • McCMcC346 Posts: 697Member
    Thanks, @psCargile !

    Texture work continues.

    Because it's come up several times when soliciting feedback, I should stress that the only (deliberate) texture on most of these is the flat color texture. I've pulled out the roughness and bump passes that I had on here before, as well as extracted out the metal color texture that I had mixed into the mess, as well. That's why the surface tends to look so much more bland than it has in prior renders; it's not indicative of a final look, but rather a stripped-down version aimed specifically at hammering out one particular kind of texture component. Even in just the color channel, I'm still planning to mix in some micro-paneling detail (see below for some experiments with this!), some procedural edge weathering, a general light scuff pass, and selectively mix how intense the actual color pass comes through.

    Finished vectorizing my sketched-in panel lines from last time (Nov 14):


    Vectorized the sketched-in texture detailing on the "notch" (Nov 15):


    Spent some time experimenting with different configurations for what I want the blue accenting on the engineering hull to look like, finally settling on the below for the dorsal side. Also vectorized the top-side pylon texture (Nov 21):


    Same, but on the ventral side of the engineering hull. This also shows some stuff I've been messing with for mixing world-projected "decals" into the texture for signage. I had previously been using a UV Project modifier and multiple UV channels to do this. What you see here is entirely in-shader, no extra UV layers required. Instead, I just use a non-rendering square plane scaled (this is important) to the desired size, and then fed through the shader via Texture Coordinate > Object, fed into a Mapping node, and then alpha-mixed ontop of the base color. There about half a dozen of these in the shader now. The banner itself is its own 4K texture, albeit not 4K on the vertical dimension (also Nov 21):


    That night, I started thinking about how I might work in some procedural micro-paneling into the hull. While aztec paneling is probably Trek's most famous hull paneling, conventional aztec was less of a thing on the filming model of Enterprise-E and was also absent on the physical and digital models of Voyager, both of which have been my aesthetic touchstones for surface detail on this ship. Instead, one particular image stuck in my mind as a fascinating inspiration: the shot of Picard, Worf, and Hawk walking across the Enterprise-E saucer in First Contact. I pulled a selection of the hull out of the image and perspective-corrected it into a rectangle and have had this setting in my reference folder for a while as something to return to.


    I spent a good chunk of yesterday messing around with how to do this procedurally. It started with a Manhattan-style voronoi texture and a bullheaded certainty that I could figure out how to create an "edge extract" effect entirely within Blender's shader nodes. Turned out that yes, indeed, I could!


    The shader node for that is pretty large, but it's also fairly straightforward. Posted here, with annotations, for anyone interested in using a similar effect:


    I took a crack at a first-pass implementation on Coro herself, with promising (if in desperate need of tweaking) results. I tried letting the voronoi component of the network use the actual UVs, which in hindsight was a mistake given the varied scale of the UVs (which the main texture UDIMs account for, but generated object coordinates would not). I also need to back off how strong the color variance is, and let the roughness -- which s not hooked up to this at all yet -- do most of the heavy lifting there. But on the whole, I was very happy with the initialy results of what started as a though experiment.


  • JESJES34 Posts: 68Member
    I've been meaning to comment for this for some time, seeing as how I love both the design and the model work. I especially love the cutouts under the nacelles, and good job filling in the rooms surrounding the bridge. Segmenting that area just was not working. And damn, that's a lot of shuttlebays, which makes her perfect for carrying fighters.

    There are a lot more images online of Mark Kingsworth Insignia class, and is hence more well-known, but among the interpretations of the next-gen design from the TNG Technical Manual, your Katana class comes off looking the sleekest. In fact, I don't think I ever even saw your design until a few years ago. As far as I knew, the only Katana class was that old design from Sci-Fi Art, which was probably made around the same time.

    Looking at all of the renders, I have thought of a few things that could help blend the saucer and engineering hull better together, such as blending together the impulse drives, engineering hull, and shuttlebays 2 and 3 together (something similar to the Sovereign Nemesis refit, where the saucer meets the engineering hull), but I don't want to suggest any unnecessary work, or overstep the boundary between helpful criticism and backseat modeling.

    Overall, she's a beautiful interpretation of an old concept.
  • McCMcC346 Posts: 697Member
    Thanks, @JES!

    Over the next several days, I spent a lot of time tweaking the various layer settings and mixes, never really finding something I was happy with. The following images are all from Nov 26th, representing different tweaks.

    coro_2020-11-26-1102.jpg&size=320 coro_2020-11-26-1742.jpg&size=320
    coro_2020-11-26-1111.jpg&size=320 coro_2020-11-26-1801.jpg&size=320 coro_2020-11-26-1925.jpg&size=320 coro_2020-11-26-1940.jpg&size=320

    One persistent issue nagged the entire time: here was this radial saucer, with panels that didn't remotely conform to how it would have been built. If only I could convert Blender's automatically-generated UV coordinates into polar coordinates, the way one might with an image in a 2D editing program...which, I realized, is just a math problem. Of course, calling it "just" a math problem was a bit naive and it took days of fiddling to get the math to work the way I wanted to, in part because I feel like the "direction" I'm running the equation is backwards compared to what it "should" be...but that's the direction to run it to make the coordinates work. In any case, after lots of fiddling, I managed to achieve procedural polar UV coordinates for the saucer on Nov 29:

    coro_2020-11-29-2035.jpg&size=320 coro_2020-11-29-1617.jpg&size=320

    I'm going to write up a fairly in-depth post in the tutorials section on how I did this, in case it's useful to others.

    Then I had to take a bit of a break from working on it, as we finally made live a fairly major system upgrade at work. I finally dove back in in the last couple of days, further refining both the panel layer mix and settings within itself (there are four distinct voronoi "mix" layers involved) and how that mixes in with the main hull shader. I also introduced a mask texture into the mix, which I can use to control the angle of the triplanar mapping for the panels on the pylons, while leaving the rest of the triplanar mapping unaffected, as well as desginate parts of the saucer to not be polar coordinates. In the prior images, I had been doing this purely based on UV coordinates (since all the saucer wedges are in UDIM 1002, so I could use the U range 2-3 as a "mask" as well). I'm pretty happy with the color mix at this point -- I think it has the right level of detail and level subtleness to achieve the effect I'm going for -- but the roughness mix still needs some work. Here's where we stand now:

    coro_2020-12-11-2147.jpg&size=320 coro_2020-12-11-2202.jpg&size=320
  • McCMcC346 Posts: 697Member
    edited January 1 #171
    Happy with the results I was getting out of my polar coordinates and micro-paneling shader networks, I couldn't help but feel like I had to finally solve the overall shader problem: how do I make this hull look good? What does "look good" even mean? I received a great deal of advice and heard from a wide variety of perspectives, but the fundamental problem facing me is that I didn't have a strong sense about my "look target" for the hull.

    In starting to explore that, though, another problem presented itself and demanded my attention first: the way I had done the saucer registry thus far -- and in previous renditions of the model, even -- just wasn't cutting it for me anymore. Squashing the name onto the same sort of plate as the registry made the name look cramped, and having them side-by-side like that bucked a pretty universally-consistent tradition throughout nearly every Starfleet ship of "big registry, smaller name above it". Even weird setups like the *First Contact* ships, with their strange placement for this text on the hull, were consistent about the rule. That signaled to me that I needed to put the registry front and center on the saucer, interfering surface detail be damned. Thus sparked a *whole* lot of fiddling with different font sizes, font kerning, placement, and so on, to finally settle on this:


    In the course of figuring that out, I also noticed that my rooms had seemingly all gone dark...and with good reason! When I arranged all of the UVs of the various objects that make up the ship into a unified set for UDIM textures, I had moved the room UVs into "negative" UV space, since they weren't part of the UDIMs and were instead handled on their own. But! The shader I had been using expected them to exist in the 0-1 UV range. A few shader tweaks later to account for their new negative coordinates, and Coro's crew were no longer lost in darkness. I also finally added the beginnings of a room-by-room randomizer to the shader that selectively disables their emission based on their UV island index, meaning the crew wouldn't always be blasted with light, either.

    coro_2020-12-19-1535.jpg coro_2020-12-20-1016.jpg

    With those issues settled, I returned my attention to the hull shader. Since Enterprise-E is probably Coro's closest relative in terms of hull patterning, I decided that that would be my "look target". I concluded that the only way I, personally, could really figure out how to translate "hull looks like Enterprise-E" into shader parameters in Blender was to choose a few screengrabs out of First Contact -- I wanted shots where the ship was depicted by the physical studio model, or else I'd just be chasing a CG target -- and try to achieve an approximate match in the way presumptive light interacted with my hull shader that aligned with the screenshot.

    Brief aside about that. In my shader exploration, I had indeed studied a number of photos of the model itself, outside in broad daylight. Trying to match that, though, would not achieve the result I was going for, because what the actual mode looks like in actual sunlight is not what the depicted spaceship looks like in depicted sunlight. All manner of studio lighting fakery, multi-pass compositing, and so on factor into how Enterprise-E looks on film where it is supposed to be lit by "sunlight", vs. how the studio model looks in a parking lot under real sunlight. Fundamental to what I was after is asserting the premise that what we see on screen, after all the movie magic has been done to the studio model, is how the ship "really" looks.

    Armed with that mission, I selected three shots: the opening reveal of Enterprise-E with the nebula behind it and two shots related to the deflector dish spacewalk. In all three, I also took as a given that the ship was being key lit by the sun (an admittedly questionable assertion for the nebula shot), allowing for some color variation -- again, I was after getting the hull to react to light as desired, not so much worried about color-matching. For the Earth-orbiting shots, this felt like a pretty solid assumption. Finally, I allowed myself to use either an environment image (specifically for the case of the nebula; I essentially took the screenshot itself and applied a sharp curve to it, then cranked up the intensity of the output to make it simulate an HDRI) or a wide-angle fill light, to account for the typical high-illumination way in which most physical model shots in *Trek* are typically lit, whether it makes physical sense in the scene or not.

    I stripped all my textures off the hull so I was playing purely* with the parameters on Blender's Principled BSDF. Here were my initial results:

    coro_2020-12-27-2023_fcneb.jpg coro_2020-12-27-2223.jpg coro_2020-12-27-2306.jpg

    * This isn't strictly true; I actually applied a very fine noise to the color, roughness, and bump channels to give the surface the faintest bit of imperfection. I would end up retaining this noise as a foundational element of the new shader.

    The biggest surprise for me was how dark I needed to make my base color: 28.9% gray. Baseline roughness sat around 38.5%, with a IOR of 1.585 (which works out to a Specular value, using the Principled specular formula of around 0.64).

    The next hurdle was figuring out the right mix for my micro-paneling detail, which was at this point also serving the role that aztecing often does. I put together a few different options, being careful to ensure I stayed roughly in that ballpark roughness range I had identified, and polled a bunch of folks on what they thought looked best.


    The end result of that feedback:


    What followed (and continues) was a whole lot of adjusting color texture colors and how those mixed in, as well as rebalancing all my existing roughness and bump maps to mix properly to follow with the look I had now pretty firmly established for myself.

    coro_2020-12-31-1144.jpg coro_2020-12-31-1734.jpg coro_2020-12-31-2124.jpg coro_2020-12-31-2130_fcneb.jpg

    The middle two images there also include some experimenting with doing a better-looking "bloom" composite pass...still very WIP, though. The final image there was done by request, and revealed several things: my key light in that setup needs to be bluer, the registry light is too bright, and I forgot to turn the nacelle grille glow on!

    On with 2021...
    Post edited by McC on
  • McCMcC346 Posts: 697Member
    I next spent some time adding some much-neglected accent lighting around the ship, and retooling the way I had been handling my running lights. In particular, I was missing the white hotspot in the middle of the otherwise colored light, so I revised how my running light object was built in order to achieve that look. I also spent some time placing decals like the ship identifier and Starfleet pennants on the flanks of the stardrive and under the nacelles.

    coro_2021-01-02-2140.jpg coro_2021-01-03-1125.jpg coro_2021-01-03-1600.jpg

    I moved on from there to finishing the last set of textured panel lines needed on the saucer wedges, which also prompted some rebalancing of my base metal "grunge" texture that feeds into the painted maps.


    At last, it was time to tackle the nacelle textures.

    coro_2021-01-10-2224.jpg coro_2021-01-16-1110.jpg

    About this time, I started to realize that my UV layout for the nacelles was garbage. Even with 4K maps, details painted into the texture were looking too low-res up close. The nacelle was also set at bit of an angle, which made all of the panel texture painting a giant pain. I set about completely re-UVing the main body of the nacelle, in the process switching it to use its own 4x2K texture instead of a 4K square texture that it shared with the Nebula-style sensor panels on the ventral saucer. With the nacelle now easier to work with, finishing it up happened fairly quickly.

    Unfortunately, I ran into a hard limit in Blender/Cycles with my shader at this point -- it had grown too deep to render anymore, and was showing up all black. I decided at this point to "bake out" a large 4K image of my micropanels, rather than rely on them being entirely procedural (which required quite a few nodes). I'd still keep the procedural conversion between triplanar and polar coordinates, but there was no need to keep the four distinct Voronoi procedural texture networks that comprised the layers of micropanels at this point -- I hadn't tweaked those particular settings in quite some time. This also gave me a chance to simplify my math a little bit, which has had a very welcome impact on shader compilation time in the viewport, and even some impact on Cycles render time.

    coro_2021-01-18-1616.jpg coro_2021-01-18-1831.jpg

    I next spent some time rebuilding the procedural shader on my warp nacelle grilles, aiming to get a better lighting result while still remaining fairly true to the Sovereign-style nacelles that are their inspiration. This involved a little bit of rethinking, switching from using stretched Checker textures to simply using some sine math on the UV map itself to determine the masked-out bands where the dark copper nacelle spacers would be and the gradual dark-to-light fade for the emission surface. I can also turn the emission of entirely, which results in a sort of flat, dull gray surface that matches what you might see of the ship when it's powered down.


    Hull shader work, it seems, is never done. As I worked on finishing the panel lines on the flanks and spine of the ship, I continued fiddling with the color, roughness, and bump mix for the micropanels. One major "advance" here was better parameterizing my polar coordinate conversion node group so that I could independently control the latitudinal and longitudinal scaling of the texture; previously, they had been directly linked, which resulted in major stretching on the rim of the saucer. I further mixed in some slight perturbation in the IOR/Specular value informed by the micropanels, still centered on an IOR of 1.585, but fluctuating between 1.5 and 1.67 now.

    coro_2021-01-24-2156.jpg coro_2021-01-24-2228.jpg

    The second image here can be directly compared with the similar image from two paragraphs back to see the changes in glossy reflections, color mix, and tiling.

    I remain unsatisfied with total shader result, though; it continues to look "too CG" to me in most renders, which is frustrating.
  • McCMcC346 Posts: 697Member
    I've spent the last several days making very little progress on textures and continuing to obsess over the hull shader. Discord quote of the year:
    lewisniven wrote:
    2071, the grandchildren play with their hoverboards in the garden as grandpa McC further tweaks his hull shader

    I keep finding things that are leftover -- or built in a way that's no longer valid based on my current approaches -- in my shader that are "corrupting" the look I expect to get out of it.

    Since I had broadly homed in on the major material properties I wanted already, I finally felt comfortable following @Viper's suggested approach of dropping the ship into a color-corrected HDRI lighting environment, rather than space lighting, and adjusting material properties from there. The following images represent that shader exploration over the course of 1/29 to 2/1, with the last one having more-or-less finally achieved a look much closer to what I intended -- thanks in part to tweaking some settings based on this approach, but also thanks in large part to tearing the shader down to its constituent pieces and adding them all back together bit by bit, ensuring that everything was combining exactly as I intended and expected.

    coro_2021-01-29-2043.jpg coro_2021-01-30-0739.jpg coro_2021-01-31-2253.jpg coro_2021-02-01-1403.jpg coro_2021-02-01-1751.jpg coro_2021-02-01-2018.jpg

    After that, I did a little rework on my textures themselves, mostly nudging colors around. In particular, I made the textured gridlines brighter than their underlying surface by a uniform amount across the whole ship, and I made the medium-gray armor panel lighter commensurate with the "off white" panel color's distance away from white, so the ratio between them is now the same. I finally shifted the escape pods away from their sandy color and to a more off-white/creamy color, as well.


    I'm fairly happy with where this is now, though there are a handful of relatively small tweaks I already see that I need to make: reduce the bump depth on the panel lines and the panel line color on the dark "cape" (and possibly on the blue panels) is a little too dim to make out properly. Still, those are fairly small changes and I feel like I can finally move forward with further texturing work now.
  • McCMcC346 Posts: 697Member
    More texture and shader work across the ship. Entering the home stretch now.

    coro_2021-02-06-1225.jpg coro_2021-02-06-2042_partial.jpg coro_2021-02-06-2209.jpg coro_2021-02-06-2221.jpg coro_2021-02-07-2036.jpg coro_2021-02-13-1726.jpg coro_2021-02-14-1824.jpg
  • McCMcC346 Posts: 697Member
    Little remained after all the stuff in the last post, but I wasn't quite done yet!

    I experimented with ventral registry, which I ultimately decided against. Dorsal regsitry's big enough to span that middle spar, but ventral isn't. Tried different placements other than centered, but didn't care for them.


    Revised my room interiors to vary what you can see through the windows -- either quarters or hallways now, instead of just quarters everywhere, depending on what part of the ship it is. Totally rebuild both the quarters textures alongside the hallway ones to provide more variety.


    Also rigged up the rooms to respond to different alert statuses, controlled by a single "alert status" parameter that drives shader changes. Yellow, red, and blue (for landing) alerts:

    coro_2021-03-09-1738.jpg coro_2021-03-09-1746.jpg coro_2021-03-09-1755.jpg

    From here, I did a whole deep dive into the branched path tracer, instead of using the regular path tracer, to finally find a good balance of render speed and noise level (or rather, denoising artifact level). After a lot of iterations and experiments with progressively more or fewer subsamples in different combinations, I settled on 90 AA, 3 diffuse, 2 glossy, 2 transmission, 2 mesh light, and one of everything else to provide a clean look I was happy with, without taking hours to render. That combination, for the render angle I was testing, only takes about 9 minutes, in fact!

    Now it was time to zero in on a final render "look". I started from enotobian's blog post on his post-processing method and did what I could to adapt it to Blender as a starting point, then iterated on it from there.

    When I was finally happy with my results...I called it done at last!

Sign In or Register to comment.