Harvestman - Thunbergia gretae

[grant.r.brown]

I have been working on making an accurate (as much as I can) model for a male of the harvestman species Thunbergia gretae (named after Greta Thunberg).

I don’t have much experience sculpting, and also no stylus, so I made this “by hand” (without sculpting or retopo). I focused on first mapping out edge loops for the various tergites and sternites (the armour plates) and then trying to reconcile everything in between. I believe everything is quads except for a few triangles where things meet at a point (e.g. the horn on the back).

It’s been quite a challenge working from line drawings and trying to translate this into 3D, and some structures were not measured (e.g. all the legs). In these cases I had to rely on photos and measurements of species that were in the same genus or family. I couldn’t find any good diagrams of the mouth structures behind the chelicerae (it’s been a while since I worked on harvestmen, and you don’t normally dissect the mouthparts) - so these are a sort of “indication” of some structures I do know are present, but are not 1:1. I do know the author of the species description, so I will ask him how close I got when I get a little further along.

I need to sort out the back end, as the anal operculum as not that accurate and is a bit of a bodge. Also tweaking some of the overall form and smoothness of the tapering. This species has various little spikes on the palps, etc., as well as some hairs (setae) which I will need to add. At that point I may be able to thin out some of the topology a little (but I don’t think I can face retopo’ing this). Then will be to rig the model - I’ll need to think about the best way to do this. I’d like to have constraints so that the legs and palps move with realistic limits on their bending.

Texturing also will be interesting - this species is mostly black with little golden spots, so it won’t be the most impossible thing to do, but I do struggle getting a realistic looking cuticle/integument texture.

This has taken me about 2 days to get to this stage, inclusive of looking for reference photos and trying to decipher diagrams, etc. I know some would sculpt this in a few hours, but hey ho. Sculpting with a mouse is horrible.

Face

Top View

Palps and chelicerae (awaiting details of spines, etc.)

Edge loops highlighted

The legs of these families of harvestmen are ridiculously long.

Without subdivision modifier, the mesh is about 12K vertices, which I think is not too bad. Obviously some areas are still heavier than others so it could be optimised. You can also see how I at one point applied the subD on the main body then decimated back, which I haven’t done on the legs or palps.

The reference line drawings. There are no photos I can find of this species.

[grant.r.brown]

I decided to apply the subdivision and then decimate>unsubdivide again, the mesh is higher poly now at around 50K but it looks decent even without any further subdivisions (whilst the previous low poly mesh ended up being about 300K verts with 2 subdivision levels). Of course, some of my quads were lost in the process so I will need to tidy these up.

The pedipalps are a bit of a nightmare - the morphology is not perfect to the diagrams, but it should be close enough from a distance.

50K verts.

An attempt at UV unwrapping too.

[grant.r.brown]

Well, it’s been a lot of faff but I have made some progress despite bug issues.

I added some further details of the characteristic denticles found on the coxae of the legs (legs 1, 3, 4) as well as using geometry nodes to scatter hair and more denticles in areas reported in the paper. I thinned the legs as well, they were too squat on reflection.

I had to fix an area by the legs as this was not that accurate for the joints. Still not 100% true to life but this seems much closer to harvestman legs now.

A little test render, just to see if there are any issues with normals, etc. I think texturing this properly will be quite a challenge. Too bad I don’t have substance painter.

Still need to create a texture map for the spines. I think the texture of this species in life is probably similar to this other Gagrellid, albeit with far less “gold dust” patches.

[etn249]

@grant.r.brown

I think texturing this properly will be quite a challenge. Too bad I don’t have substance painter.

Did you have any plan as to how you would do the texture?

I think I have an idea how I would do it. I find Blender’s painting is too glitchy and too low performance to directly paint something this detailed, but it’s good enough for creating texture masks. I would make multiple material trees (multiple principled bsdfs inside the same material), each with its own set of textures. One material tree for the black areas, one for the gold and one for the legs, each with its own textures and material settings (ex. the legs’ material could be slightly translucent like on the photo).

Then, I would paint masks to affect which material appears where on the model. The advantage of masks is that they can have a lower resolution than the textures they affect: you paint on a relatively low resolution mask and get a high detail result. You get to literally paint with entire materials. Also, the fact they aren’t used directly as the texture itself means the glitchiness of Blender’s paint mode gets somewhat hidden.

It’s also possible to add more masks for different purposes, like one could be created to affect the amount of bump, so the smooth areas can be created.

Of course, at the end you get a pretty complex material, but it should be possible to bake it to a single set of textures.

[silex (Jakub Jaszewski)]

Nice progress with the geo-nodes!
The texture on the thorax and the abdomen can be done with proximity attributes. You can find a way to separate specific mesh islands, then sample distance from the edge and pipe it to shader editor.
This attribute can also be used to scatter some tertiary details on the mesh inside geo-nodes where yellow=more/bigger details and black=none/smaller. Same for displacement.

[grant.r.brown]

@etn249

I think I have an idea how I would do it. I find Blender’s painting is too glitchy and too low performance to directly paint something this detailed, but it’s good enough for creating texture masks.

Yes, this is how I am approaching this - it’s about my limit for texture painting to just do b&w masks. I do I had a tablet though, using a mouse is a grind. What sucks about not having a job at the moment is not having access to things like this or Photoshop, etc. Although a bonus is that I am finding there are some decent online alternatives (Photopea.com is not too bad actually).

@silex

The texture on the thorax and the abdomen can be done with proximity attributes. You can find a way to separate specific mesh islands, then sample distance from the edge and pipe it to shader editor.
This attribute can also be used to scatter some tertiary details on the mesh inside geo-nodes where yellow=more/bigger details and black=none/smaller. Same for displacement.

Yes, I have been using your suggestion on my other work too (Manpower) for edge wear, it’s very powerful and way better than trying to muck about with bevel nodes in the shader node tree. In the species description it does indicate where these gold spots are (the species in the photo is waaaay more golden than T.gretae it seems). So I think for these I will just make a simple mask from the diagrams. However, I like the idea of using these attributes/fields to control instance scales etc.

[grant.r.brown]

I’ve been working on two different aspects of this project - texturing the harvestman, and trying to make geometry nodes moss.

I was not getting on all that well (keep hitting walls), so I decided to “produce” something to keep my motivation up. One thing I wanted to do since the start of this project is to create a render that looks like a macro photograph. I have an idea for what this will look like, but I made a quick “half scene” in the last hour or so at real life scale to see how this might work. The harvestmen is 6.6mm in body size, which is tiny, and the gingko leaf is within the range of true leaf sizes for comparison. It’s interesting trying to render at this scale, the camera in Blender doesn’t quite behave like real macro lenses. It’s all a bit academic anyway, as in final scenes I will scale things up so this detail is actually visible.

I need to finish modelling the procedural star moss and also properly model some dead gingko leaves (the species is found in China, and gingko is one of my favourite trees). If I had money to burn I’d buy something like this and just remove the leaves. Oh well, no money to be lazy.

With the nodes I need to figure out how to introduce stem flex and also how to grab the attribute that would let me set up the shader to colour the leaves brown near the bottom (dead ones). As usual, simple things with GN seem to be a total pain in the arse.

Other jobs to finish:

I need to figure out how to make the gold patches look better. I tried to use real geometry with geonodes to give them a powdery look, but I can’t seem to get a good look, and it’s also insanely heavy to render.

The cuticle is a mix of different textures to give a sort of rough appearance, but with the cells visible:

Similar to this:

Still need to do various things (e.g. sort out leg details, etc.). I can’t believe that you can’t export UV island outlines from Blender, what a faff.

I’m glad this species is basically entirely black however, I don’t know if I could replicate some of the more detailed species.

[silex (Jakub Jaszewski)]

I really like that you are putting so much attention to detail with this project.

About gold patches with geonodes. If your instanced geometry is too heavy, you can try to substitute it with a simple cube then increase density. This paradoxically can reduce final polycount during rendering.
If you substitute your model with a cube you can easily break its sharp edges with a bevel shader node.

I don’t know how usefull this might be, but if you want to displace stem randomly you can do it with very simple displacement setup:

To pin stems to the ground you will need falloff, so check @higgsas node pack here on the forum.

Alternatively you can check following tutorial. It’s different technique, but you can still utilize parts of it:

Blender Tutorial: Noise Advection with Geometry Nodes

[grant.r.brown]

Some different assets - Gingko leaves.

I finally have a working copy of Materialize (on my old pc it just never worked no matter what I tried), so I made some PBR textures from photos of gingko leaves (Freepik, etc.) and made detailed leaves suitable for macro style shots. They are quite easy to make - draw vertices around the shape of the leave outline, delete the outer faces, then retopo with quad remesher. I sculpted some details and generally shaped the stem and leaf to try to make them look less like planes. The only bottleneck is finding good images of leaves. When I used to live in the UK I actually had a ginkgo tree I could have photographed, but that was sadly left when I moved. Everything looks a bit dry at the moment too, so I might try to make it look a bit more damp.

Viewport (need to replace the rock asset, sort out ground plane, add other stuff, etc.):

I downloaded Blender 3.3 and also tried the new hair curves system to see if this would be better for the setae and spines of the harvestmen. I would say I didn’t notice a significant improvement to performance, but the ability to comb and shape hairs is pretty cool. For now I’ll probably stick to Blender 3.2 however for this scene, just because I can’t be bothered setting up all my addons and defaults again. .

I’m going to see if I can make better cuticle textures now with Materialize, perhaps I won’t need the spines and such anyway. Even at close range they are quite subtle.

I suspect the same could be achieved with textures for the gold dusty patches:

These are heavy enough in terms of geometry that I can’t actually viewport preview both of these systems at the same time, so I’d like to make the model a bit lighter.

When it comes to renders, for these shots I want to emulate the Canon EF 100mm F/2.8-32 USM lens. I used to have one of these and it was pretty great at photographing insects and spiders. I also used to have access to the Canon MPE-65 lens, which was just crazy to use. However at a max f/16 it was hard to see much and the zoom functions were a bit pointless as you needed so much light you fried anything you looked at.

[grant.r.brown]

A procedural moss stem, based on Bryum pseudotriquetrum (I mean, sort of).

I made the leaves the same way as before - they are quite high poly, but given how close the shots will be it seems required. I then made a geometry nodes set up to generate random stems:

The material is from Materialize with some tweaks to SSS and transparency, etc.

[grant.r.brown]

A test render, will likely redo the composition.

Need to finish texturing the harvestman properly, and a few things to sort out - textures for rocks, etc.

Happy with Ginkgo leaves. I think that my moss doesn’t stick out as obviously being made by me compared to some published moss assets, so I am happy enough with these. Shame the rain drops didn’t really show that well on the leaves.

[grant.r.brown]

I decided to do away the mesh tubercles and instead made a texture to indicate these. Performance is way better. I had to redo the shader tree, it makes a lot more sense now, although I don’t think you are supposed to join more than two normal maps together (it sort of works so good enough).

I also moved to 3.3 and made use of the hair curves to create the setae on the palps; the other spines on the legs are normal geometry nodes instancing curves.

I think I need to tweak the roughness, it looks a bit like polished leather just now. If I can get the displacement to behave then I might be able to get those crusty gold-yellow patches to look better.

I need to add some dust and dirt to the specimen too, it looks a bit too clean compared to most photos of Gagrellids. They do like to clean themselves, but there’s always bits of dust, pollen, etc. on them.

[grant.r.brown]

Finally made progress with the dusty/powdery patches using displacement:

The current tree also emphasises the tubercles on the legs and the cuticle texture, without ruining the integument.

At render resolution (adaptive subdivision)

Think it’s time to start making the final scene composition and then doing some portrait renders.

[joseph (J)]

@grant.r.brown

I can’t believe that you can’t export UV island outlines from Blender, what a faff.

Extracting a UV layout

[grant.r.brown]

Yes, I know how to do this, but this is not quite what I mean. What I wanted at that time was just the island outlines - I thought you might be able to do this by hiding all edges that were not seams, but sadly not. Setting the opacity to 0 also does not work sadly.