logo80lv
Articlesclick_arrow
Research
Talentsclick_arrow
Events
Workshops
Aboutclick_arrow
profile_loginLogIn

Breakdown: Creating Realistic Fan Palm Using SpeedTree & Substance 3D Painter

Gerard Ivanov discussed his work on the Fan Palm project he made using SpeedTree and Substance 3D Painter, showing how he managed to create lifelike leaves in just 24 hours.

Introduction 

Hello, my name's Gerard Ivanov. From a young age, I was an avid gamer and had a huge interest in game development. 

Initially, I had an interest in 2D illustration, but then switched to graphics design. At some point, for my graphics design projects, I needed simple 3D graphics, which turned out to be my gateway into 3D. Gradually, I learned 3D starting from 3ds Max and then switched to Modo, which is my core modeling tool to this day. Initially, I started with hard surface modeling, focusing on weapons, vehicles, and military gear. 

I was self-taught with great help from Ilya Makarov and Dmitry Bely. Since 2021, when I joined the studio where I work right now, I've contributed to several Call of Duty titles.

Fan Palm

After my recent vacation, I got obsessed with palms, so I decided I needed to make one. 

Since I have a shortage of free time due to working on a personal game, I had a strict time frame for this asset – 24 hours. So every minute counted, and I had to cut corners and consider how I allocate my time well. 

My approach with references is to keep it simple since every plant, even within the same species, can greatly vary depending on the environment it grows in. Scraping everything there is on the internet is something I avoid. Usually, I aim for 3-5 core references similar to each other that I will be working with and then support them with detailed references of parts of the plant, like stem, bark, leaf, and so on. Essentially, this is what my core references looked like:

The initial draft of the asset was done entirely in SpeedTree using whatever suited me on Megascans. I consider the initial draft to be the most important part of the work because it's the groundwork for the big picture read and gives an understanding of how hit-or-miss I'm with the references. Everything that comes next is built upon this foundation. The initial draft took me an hour using Megascans for the trunk and leaf atlases. The end result of my one-hour draft looks like this.

And here’s a GIF of how it was built and what materials were used for each stage of the palm’s draft generation logic.

Here is the video tutorial if you are interested in making your own fan palm draft. But first, we will need the following materials from Megascans: trunk scan, stem kit, and leaves.

From this point, I was happy with how my draft looked, and I replaced everything except for the Megascans trunk base and trunk tileable. Now I needed to set up my own material and assets for everything else.

I began with modeling separate leaf meshes, frond meshes, and frond “teeth”. I kept them really simple because I would be overpainting my final bake anyway. And also, I was on the clock with this asset!

After making my kit, I switched back to SpeedTree to set up my atlas scene. This is how the end packing looked. For this asset’s atlas, I used only half of the texture space because I didn't feel the need for more elements. This approach is often used in production to save memory. 

For the green leaves, I decided to make them beefy tris-wise and set them up as separate leaves. This way, I could make really cool close-ups and model more complex geometry for every leaf for a “dripping” effect instead of going for huge cards. But huge cards, like for the dry ones, are used because it’s way cheaper tris-wise. I didn’t really aim for some kind of strict tris budget, so I went crazy with green leaves.

Dry leaves, on the other hand, went as single “packs” because they bunch up on the trunk, and it’s harder to notice exactly what is going on there. The same goes for dry fronds, which are used as a “connector” for dry leaves and as separate pieces of dead frond leftovers.

The stripe on the right is not the card (two-sided plane). These frond (branch) pieces are used as closed geometry because it’s a noticeable element of the palm, and I decided not to use a card approach for it.

The “teeth” are just a card trim attached to green fronds. Here’s a breakdown of the elements:

After my bake was done, I spent a limited amount of time (which was running out!) doing additional texturing inside Substance 3D Painter. I added some gradient details and damage to the green leaves.

Now I needed to update the cards. Some cards for this I made with Modo and some inside SpeedTree. There’s nuance when to use one approach or another, but often they are interchangeable. For example, I needed pre-bent leaf cards, which are easier to achieve by modeling a card outside SpeedTree. Cards for dry fronds, which are to be simply bent down, are totally okay to do inside SpeedTree.

So, now I replaced everything, except for the trunk, with my content. I ended up with 3 texture sets: 1 for trunk scan base, 1 for trunk tileable, and 1 for everything else.

The only major change, compared to the draft, is how I approached the green leaf cards. Instead of a single huge leaf card, now I had a single card per leaf – expensive, but looks great!

And we are done!

My rendering setup was very simple. Since it’s a “speedrun” (with SpeedTree, haha), I used SpeedTree’s rendering viewport. My only post-production was to sharpen it up a bit and balance out highlights and shadows in Photoshop.

Conclusion

My biggest challenge with this asset was to fit it into a 24-hour timeframe. I had to allocate time for each aspect of the work very carefully.

My advice to beginners in foliage is to learn to draft quickly and close to the reference. I do this practice all the time when I pick a limited number of references and draft out as fast as I can, and I often scrap the entire work for a new iteration. The biggest noob trap here is to cling to initially unsuccessful drafts. In my practice, a bad foundation is a bad result, no matter how much more time you dump into it. Build simple and iterate carefully. Ask feedback from as many people as you can, even if they don’t model! Humans have a very good eye for anything that seems wrong with foliage, and that can be a huge help for you.

Gerard Ivanov, Vegetation Artist

Interview conducted by Gloria Levine

Join discussion

Comments 2

  • Anonymous user

    Make 102 dalmatians in blender please

    0

    Anonymous user

    ·21 days ago·
  • Ponomarev Anatoly

    What a legend

    0

    Ponomarev Anatoly

    ·21 days ago·

You might also like

We need your consent

We use cookies on this website to make your browsing experience better. By using the site you agree to our use of cookies.Learn more