42 replies ·
posted April 6, 2021 7:59pm
This is my project thread for the April 2021 collaboration.
I like where that was going, but I ran into some problems with it.
It was really tricky for me to get the leave alignment right with the sapling tree generator.
Also, when I zoomed in on the project image the leaves aren't maple leaves, they're just hexagonal leaves. I should honor that, and it'll save a fair number of vertexes.
I'm going to play with the geometry nodes and see if I can't come up with something I like better.
i really like the tree , those leaves are looking amazing too , nice work
Thank you yyukinoh1989!
Thank you yyukinoh1989.
As much as I like what was going on with the sapling tree generator, it's a little too random.
Compare the sapling generated tree on the left to ones on the right.
The ones on the right are hand modeled with the skin modifier. They're much easier to control.
Now I just need to get leaves applied. (Not sure if I'll need a lot of smaller branches or not.)
wardred I'd say that (at a guess) the trees on the right wouldn't really need any more branches because of how far away they are from the camera, but check in with you Team Leader to be sure about that hypothesis.
Other than that it's all looking good to me. 👍
wow i would not have thinked about modeling a tree trough a skin modifier , great idea , and its paying off :)
yyukinoh1989 Thank you, I'm pretty happy with the results.
Somewhere out there is a tutorial I think Kent did on creating a spooky tree that went over creating the tree with the skin modifier, and a single subsurf modifier to make it a bit rounded.
The only downside is once the modifiers are applied the branches alone are 80k vertexes! It's relatively easy to slide the redundant loops into each other then merge by distance, but that'll take me a minute to do.
Still exploring the best way to get leaves onto these beasts.
vincav81 - I need to go back to the project guidelines a little. You're correct, from the perspective of the photo, I could probably reduce the number of limbs significantly and have leaves floating with no attachment to branches at all, and one wouldn't be able to tell because they're so far away.
If the goal is to be able to "fly" through the model, and be able to take good photos of it from other perspectives, it's possible one of the mid ground trees could become in focus, in which case having leaves obviously floating away from the limbs would be problematic.
Just messing with floaty leaves. I actually gave the stem some depth. . . but I think having a stem may have been a mistake.
I like the trunk form the skin modifier more, but I actually like the shadows from the sapling generator more.
Man, sapling generator, particles and the hair generator, geometry nodes. . . I never realized their were so many ways to try to model a tree.
Edit: On the last photo the trees in question are on the right.
Homework Submission - Week 1 - MG Trees
theluthier - Finally got to a spot where I could link you directly to this homework assignment.
The trees in the first image on the right are the most complete. I really like their branching. I'm still not quite as happy as I could be with the leaves themselves.
The tree in the middle with the huge maple leaf under it was the sapling generated tree. I like the shadows from that tree better than what's happening with the leaves on the right.
In the first image on the far right and left are some skeletal trees. Not sure if I should actually be doing anything with those or not, but it kinda looks like they'd be there from the reference photo.
I asked a question about the second image. I wasn't sure if I should be trying to recreate something that would look good if you walked up to it, which would end up being the middle tree, or if I should focus on just what one can see from the photo, like the two outer trees.
Hi wardred Seems you might've forgotten to tag Kent with "@theluthier" under the homework submission title to send him a notification. Otherwise, great job on the trees! 👍
aartifact - Thanks for pointing that out. I emailed Kent earlier while I was at work, and finally got to a spot where I could actually add him to the specific homework post.
The dangers of turning things in so close to Midnight!
Wonderful week 1 work wardred! Very promising to see how far you've come already. You've earned full points from me 👍
I stand beside the same feedback I gave you elsewhere about removing the small branches for MG trees. It's worth saving the geometry the further we get from the foreground zone. Therefore these MG trees will lean on the leaf volume hiding the missing small branches.
I want to offer 2 stylistic notes: The leafy feel good and "leafy"; clearly individual leaves are in use to good effect. However right now they feel a bit sparse and lacking a more solid core. I've been recommending foliage artists try the approach from this video:
It's not created with Blender and it's a bush instead of a tree, but the concept is what's important: A solid chunk of geometry representing the "core" of a leaf volume with a layer of alpha-mapped leaf cards making up the outter surface. When done well it's an effective illusion.
I agree. The shell I created. . . feels like a shell. There isn't enough interior volume which makes it look like a balloon with a loose covering of leaves and doesn't create any interesting shadows.
Leaves are definitely something I'll be working on for this week.
Homework Submission - Week 2 - MG Trees
theluthier - Some progress.
I haven't made as much progress as I wanted this week. I still have more trouble than I like aligning things. I need to spend more time with the tutorials.
I'm loading what I have to my WIP file on Google Drive.
I wasn't able to check out the file in the drive since it looks like it relies on some linked data that isn't there:
I'll check it out and give a grade for week 2 as soon as it's ready!
Thanks for checking. I've corrected the broken links.
I was working on the leaves, but I have a ways to go on them.
Once I figure out their orientation, I have a fair amount of texturing to do to them.
Edit: I believe I've fixed the broken links.
Apologies for the late feedback wardred
I can see in your file that you're doing a lot to stay true to the artwork. This is admirable but you can relax this commitment when it comes to the MG trees. The idea is that I will be able to duplicate your tree around to several locations. Which means it shouldn't be to matchy matchy to a certain tree. Just get close-ish to the shape, within the spirit of the art, and you'll be golden👌
Also Lampel has run a test and confirmed that the alpha-mapped approach to leaves is far more efficient than the individual particles. Check out the livestream from Thursday (starting around 50 min) for a demo of this workflow. It's for a bush in the stream but the exact same workflow applies to tree leaves as well.
theluthier - Thank you for the feedback!
I miss Thursday's streams. . . they're not at a good time for me, but I was excited to see the bush part of it. I'll be going through that this evening.
I need to get hopping on that as I have to get that done, and hopefully some texturing too.
Homework Submission - Week 3 - MG Trees
jlampel - Using Kent's Thursday stream, I'm pretty happy with how my tree turned out. I tried using the second set of gradients plugged into the emission. . . but I didn't like the result as much, so it's not plugged in.
One thing I noticed about this method is how good it looks is highly dependent on the lighting of the scene.
Let me know if I should put together a slightly different shader or shaders for the trees on the left. The "lower" grouping of them are all red. The upper grouping is sort of a washed out blue/grey/green.
Any other feedback you have I'll gladly accept. The bark, for instance, has no texture and is just a solid brown color. Matches the artwork, won't really work that well for a walk around.
Let me know if I did any bone headed things like neglected to link everything you need access to. My leaf texture is in the appropriate place, and I tried checking the 'automatically pack' option, but there are so many of us if there's something I missed and can fix that will make it easier for you, please let me know.
Nice work, this is looking really good! Definitely your best version yet.
I think, when it comes to the shading on the leaves, there's a sweet spot for the number of faces. I tried deleting 75% of those and I think it actually works better because each face isn't getting cut off by other faces as much:
Of course it removes some volume that would have to be replaced, but you get the idea. For the leaf cards, try to use as few right angles as you can so there's less obvious straight lines:
Some translucency also really helps here so that the one card isn't way darker than the card next to it that's facing a different direction:
I would also prefer to see a bit smaller leaves (compensated for by a bigger canopy) so that they better match the scale of the other plants in the scene.
The trunk could use some texturing as well, even if it's low resolution and really simple.
Lastly, using alpha clipped in the material should be slightly faster to render than alpha hashed since only one sample is needed.
Full points for week 3. Hope this helps and keep up the good work!
jlampel what is the difference between alpha clipped and alpha hashed?
If I'm reading it right, computationally alpha clipped is a simple yes/no. Alpha hashed does some statistical calculations which increases the computational load. For something as noisy as leaves, particularly where it's just a black and white silhouette of a leaf in the background, that extra computational load probably isn't buying one a lot.
That's exactly it! Alpha hashed is really useful for things that are partially transparent since it tests the transparency of every pixel several times and averages out the result (not unlike cycles averaging out the color of a pixel from each sample), but since the leaves are either full opaque or fully transparent, alpha clipped lets us get the same result more efficiently.
That's a great explanation of why one might want the more expensive computation. I hadn't even really thought of partially transparent objects!
Thank you guys :D
wardred, it looks like you are finishing up with your project. And, your tree looks great. I'd love to post your artwork in the TSMF Blog. Please let me know if you create a money shot.
I'd be honored to have it featured in the TSMF Blog. I have a bunch of last minute stuff to do on Sunday. Don't think I'll have a separate stand alone beauty tree by then, but I plan to finish off my submission for the team.
Homework Submission - Week 4 - MG Trees
It waits for no man. . . and I ran out. I didn't do most of the modifications to the canopy that I wanted. I believe I made it alpha clipped, but that was it.
I got a normal and cavity map for the tree. . . but I did not do the bark shader. I think that was Kent.
Ok, no worries! So do you want to submit what you have for week 4 and wrap it up, or do you want me to add the grade after you've finished whatever changes you'd like to make?
Thank you for the thought, but I don't think I'm going to get much updated before the weekend. I've uploaded WIP_nature_MG-trees_wardred_00.blend. That neat procedural texture for the bark is Kent's, everything else should be mine.
One thing I need to do better at is having a scratch file where I do a lot of experimenting and keep the production WIPs fairly clean. . . and versioned. I like being able to go back to before modifiers are applied if needed.
wardred, you posted a screen shot of your tree in the HQ forum. Is that what you want me to use as a money shot in the blog? Can you please clarify? I have already removed the Blender grid in the background and put the tree in front of a solid grey. Please advise if you want me to use it.
splat21 - Wow, that was fast. Sorry, give me just a minute. I may have a better angle soon.
wardred - the ideal resolution, due to the frame, is 1650x950 (letterbox aspect ratio). If you could put a pleasant background behind your tree with this resolution, that would be helpful.
splat21 - Sorry for any confusion.
I think these two look better:
And with just a gray background
splat21 - Just saw your note about the aspect ratio, I'll do one more round.
splat21 - Last one.
wardred - That works! And, it looks great. Congratulations on completing the project.
wardred I think this turned out really well! Congrats on finishing and full points for the last week :)
Thank you jlampel !
I love the way your tree turned out wardred! It was exactly the kind of lo-resolution, optimized tree model I wanted for the mid-ground area. I like it so much I duplicated and slightly modified a second version for some variety.
Thanks for contributing to the Collab! I hope you enjoyed and learned from the experience. It's been a pleasure working with you 🤝