Once I moved the dinosaurs over to micro voxels, the old character suddenly started to look a bit ridiculous next to them. The original player was only made up of around 12 large voxel blocks, so it was very simple and very blocky.

I actually quite liked that simplicity in a strange way. It had a bit of charm to it. But with the rest of the game becoming more detailed, it was starting to stand out too much.

So I’ve now started bringing some micro voxel detail into the character as well.

The character model is more defined than before, and I’ve also changed the way heads work. There’s still only one head available at the moment, but the system now supports swapping them out. Long term, I want players to be able to choose from a range of different heads so characters can feel a bit more individual.

The bigger change, though, is actually the clothing and armour.

Previously, most clothing was basically just a texture placed over the character. It worked, but it didn’t really give any sense of physical clothing or equipment.

Now the clothing is built out as actual voxel geometry around the player. So things like armour, coats and other wearables have proper shape and volume rather than just being painted onto the body.

I’ve tried to keep that reasonably lightweight as well. At the moment, the player character is around 15,000 triangles at the high end, which I think is a fairly sensible place to be.

Performance is something I’m always keeping in mind, especially because I’m still targeting Steam Deck as a proper platform rather than treating it as an afterthought.

The character still looks a bit funny, and I think that’s okay.

I’m not trying to make Evolve look super realistic. It’s still a voxel game, and I want it to keep some of that character. The goal was really just to make the player feel like they belong in the same world as everything else, and I think it’s getting much closer now.

I’ve also done some work on how equipment stats are presented.

All of the clothing and armour already had values for things like armour, thermal insulation and rain protection, but previously I wasn’t really exposing that information to the player.

Now when you hover over an item in your inventory, you can actually see those values.

I’ve also added a lot more information to the character screen itself. There’s definitely quite a lot going on in there now, but it means you can properly see things like your total armour, how much thermal protection you currently have, and how well protected you are from rain.

That side of the UI will probably keep evolving, but it already makes the equipment system feel much more useful because you can actually understand what the things you’re wearing are doing.

Truth be told, I wasn’t planning to touch this at this stage either, but the more I played with what I had, the more basic it started to feel.

I’d also been playing quite a lot of Valheim, and looking at how flexible their building system is made me a little bit jealous of what they had achieved. So I decided it was probably time to push mine further.

The core building blocks in Evolve are 0.5 metre voxels, which I really like because they give a nice level of detail. The downside is that placing everything one half metre block at a time can become quite cumbersome very quickly.

So one of the first things I changed was the way blocks can actually be placed. You can now drag out rows of blocks, and I’ve also added patterns through a new radial building menu.

For example, instead of placing every voxel individually, you can select a wall pattern made up of around 20 voxels and place a much larger section in one go. It keeps the underlying voxel system, but makes actually building something much faster.

I’ve also introduced what I’m calling pieces.

Pieces are things like posts, beams, stairs and smaller structural details. These sit outside of the normal voxel grid and can be positioned with much finer precision.

That means you can move a beam or post around within the space of a single voxel rather than everything being locked to half metre increments. It opens up a lot more possibilities for making buildings actually look interesting instead of everything feeling like it has been made from the same blocks.

I’ve also spent quite a bit of time improving alignment.

Previously, you had to look quite precisely at where you wanted something to go, and once I added pieces that became even more awkward. I now have a magnetic snapping system where nearby connection points are detected and pieces naturally snap into the position you probably intended.

It’s still something I’ll keep improving over time, but even at this stage it feels much nicer to use.

The radial menu has made quite a big difference as well. There are multiple pages for different building options, patterns and pieces, and I’ve tried to make sure it feels just as natural on a controller as it does with mouse and keyboard.

There’s still plenty I want to add to building, but it already feels much less like placing individual voxels and much more like an actual building system.

I actually wasn’t planning to start working on this just yet, but the prototype dinosaurs I had in the game were starting to stand out for the wrong reasons. They were basically made from fairly large blocks, so a whole dinosaur might only be made up of 20 or so visible blocks.

I always had it in my head that micro voxels would be the way to go eventually. The engine already supports them, and things like weapons and items are already built using micro voxels, so I decided to start moving the dinosaurs over as well.

At the moment I only have 11 dinosaurs in the game, but they’ve now all been rebuilt using micro voxels, and the difference is pretty huge. They still fit the voxel style of the game, but there’s a lot more shape and detail to them now.

Because of that extra detail, I’ve also had to start thinking more seriously about LODs. A Brachiosaurus, for example, is around 50,000 triangles at full detail, so over distance the models reduce in complexity to keep things sensible performance wise.

I’ve also been doing quite a bit of work on how the dinosaurs actually move.

One of the things I really didn’t like before was how they changed direction. If a dinosaur decided to attack something or turn around, the movement could feel really snappy. I had some smoothing in place, but it still looked wrong.

The new pathing works much more around proper curves and turning arcs. So a T-Rex, for example, can’t just turn on the spot. It has a much larger turning radius and has to properly move around to face something.

It sounds like quite a small thing, but it makes the dinosaurs feel a lot more physical and a lot less like AI agents just moving between points.

There’s still a lot of work to do on the AI side, and I’m definitely not where I want to be with it yet, but it’s starting to come together.

The other big thing I wanted to fix was foot sliding. Previously, the relationship between movement speed and animation just wasn’t quite right, so dinosaurs could look like they were skating across the terrain.

That’s now been fixed as well, and they feel much more planted to the ground.

Between the extra detail, the better turning behaviour and fixing the movement, the dinosaurs are finally starting to feel much closer to what I had in my head when I first started building them.

Water already existed in the previous version of the terrain, but moving from blocky terrain to smooth terrain created quite a few new problems.

The biggest one was that the water simulation itself is still voxel based. I wanted to keep that because it means water can actually react to changes in the world. So if you dig a trench away from a river, for example, the water can follow that new path rather than just being a flat visual effect sitting on top of the terrain.

The problem was getting voxel based water to actually sit nicely against smooth terrain.

With blocky terrain, that was much easier because everything naturally lined up. Once the terrain became smooth, the edges of rivers and lakes suddenly had to match much more organic shapes, and getting that to look right took quite a bit of work.

I’ve spent a lot of time improving that side of things, especially around river edges and how the water meets the terrain.

Underwater has had quite a bit of work as well and now feels a lot better than it did before. I’ve also been improving waterfalls. They’re still fairly limited at the moment, but they’re definitely heading in the right direction and I’ll keep working on them.

The important part for me is that the water simulation is still completely voxel based underneath.

So even though the terrain and river edges now look much smoother, the water can still behave like part of the world rather than just being decoration. You can change the terrain around it, create new routes for it to flow through, and eventually do much more interesting things with how water moves around the environment.

The main reason for it was actually pretty simple. During playtesting, I found that dinosaurs just didn’t path particularly well across blocky terrain. Small steps and sharp changes in height could cause problems, especially with the larger dinosaurs, and they could end up getting stuck on things that really shouldn’t have been an issue.

Moving to smooth terrain has helped a lot with that. Dinosaurs can move around the world much more naturally now, larger creatures have an easier time getting around, and overall it just feels better when you’re watching them move through the environment.

Visually, I think it’s been a nice upgrade as well. You can see dinosaurs much more clearly across hills and valleys, and the terrain feels a lot more natural.

The interesting part is that underneath, the terrain is still completely voxel based.

That means I haven’t lost the ability to actually modify the world. You can still dig into the ground, carve into hillsides and make your own caves. The difference now is that instead of everything looking like blocks have been removed, the terrain can form much smoother and more natural shapes.

I’ve also added sub-voxel precision, which gives much finer control when modifying terrain and means everything doesn’t have to line up perfectly to the voxel grid.

So although the terrain looks less traditionally voxel based now, all of that voxel data is still there underneath. For me, it’s ended up being a really nice middle ground. I get smoother terrain that works much better for dinosaurs, but I still keep the freedom to let players dig into and shape the world.

The first strike system measured reach from origin to origin, which for a voxel animal is the ground under it. Standing in front of the mouth did not matter. Standing on its toes did. And the mouth never opened. The jaw was scenery on a body that already had an attack pose.

The bite now lives on the jaw. At the moment of impact the hit is tested against that part of the rig, not the model centre, so the dangerous end is actually the face. The animation is a two-phase snap: the mouth opens through the windup, then slams shut on the tick that does the damage. Range and standoff are per species, so a T-rex has to get its head on you rather than tagging you with its ankles from a body-length away. While it hunts it can look at what it is chasing, which sounds small until you have watched one bite the floor.

Other animals already strike with different parts. A tail swing is a different problem. For a predator that is supposed to bite, the mouth had to do the work.

Evolve needs a world that looks worth crossing and that you can actually cross. Same constraint for the dinosaurs. If a player can only get over a slope by hopping every half metre, a herd cannot path it at all. A one metre riser is a step. Two metres is a wall. This slope is the worst of both. It never becomes a cliff you go around, and it never becomes ground you just walk. It is stairs, all the way down, with a tuft of grass on every tread.

That is the balance I have been chasing in the generator. Dramatic massing is fine. A mountain you remember is fine. What I cannot have is country that is interesting in a screenshot and unusable as a place to live. Plains need to be places you enjoy being in. Ranges need valleys between them that you can follow. Cliffs should read as cliffs, not as every metre of the map turning into a staircase.

There has already been a pass at that. Valley floors. A limit on how much of a map a single massif is allowed to eat. Checks at world creation that use the same walk and roam tests the player and the creatures actually run. A slice of land that stays on the finer half-metre ground so it is not all one-metre terraces. On paper, the amount of country a dinosaur can roam got better. This shot is the leftover question, and it is a fair one. The landscape can be working in the large and still be wrong in places like this.

Next is not more features on top. It is teaching the generator when a slope should flatten into something you walk, and when it should commit to a real face and let you go around.

What's in the world now is that network: long water cutting with the land, falling where the ground steps down, and meeting the ocean instead of stopping inland. Rivers show on the world map too, so finding one is a navigation problem rather than a lucky walk.

The water itself moves. Step in and you drift downstream with the current. Loose objects are meant to take that same ride: a log should be able to find the sea on its own. That's the groundwork for rafts and boats, not a feature that exists yet.

The waterfalls are already real as geometry. They still look like stacked cubes, though. The falling-water shader isn't doing its job yet, and that's the next question rather than a finished look.