Side Project: Graveyard

I am going to work on a small side project after a mobile game that I am working on has been released (mid June).

It will be a small graveyard, rendered in a game engine.

I created some documents to help me stay organised. They outline the purpose of this project, a loose deadline, the software I will be using and the approach I will be taking.

This is a moodboard I created for the scene. I'll be using this for inspiration. 

I also collected some reference images of some of the assets I'll be working on. Before I start working with the software, I will be making some rough sketches and planning out how everything should look and where the objects will be placed. 

Update 16/05/16

I made a rough sketch of where the objects would be placed. Although I'll be adding a path and perhaps some particles to draw the players eye, I have not taken level design into account during this project as I am mainly focusing on modelling, texturing and some engine work. But Once this is done there's nothing stopping me from using the assets for a level design study later on. 

Now that I have an idea of where the objects would be placed, I created a white box level to get a sense of space and scale. I spaced things out a little to and made note of the scales of each object once they felt right. I will use these scales when creating the models. I added this data to my asset list.

Update later on 16/05/16

I found this image and I like the leaf detail on the grave stone. I had time to chip away at something small.  

I decided to add some detail using English Ivy leafs as they symbolise strong family ties and loyalty according to an article I read. As I knew I'd be using the leaf several times, I sculpted it once in Zbrush. 

Then I used the MRGBZ_Grabber tool in Zbrush to grab an alpha image of my sculpt.

 

Using this alpha, I can pull out the leaf as many times as I want and won't have to sculpt it again. I can add more detail to each individual leaf later if I wanted to. 

Update 21/5/16

I made a start on the modelling of the tomb stones. 

 Made this low poly model using the scales from my white box level. I extruded the plate on the front, but looking at the results of the final bake, I could have just added it to the high poly sculpt and have being part of the baked maps.

 Exported it into Zbrush, Zremeshed it first then subdivided to about 3/4 million poly to be able to scupt really fine detail (made a new leaf btw, the other was to star like)

At first I actually retopologised the high poly sculpt which was crazy and a waste of time. I could have just used the original low poly mesh I created.. which I did in the end. 

So here we have the high poly detail baked onto the low poly mesh. Chucked a texture on to test it but will be making my own later on. 

Update 25/05/16

Spent some time on another Tombstone today. I've save the low poly versions of both so I can just go back to them, add different details and save out as new Tombstones. If I made 3 base models, I could export them into Zbrush, add a load of detail, and back them back on to the basic model, then rinse and repeat. 

I took my base cross and added some detail such as some extra edgeloops and some extrusions (as shown in the Zbrush image below)  

I masked created a mask around the areas I wanted to add the patterns to and inverted it. 

I search for "Gothic Floral" patterns and I liked the look of this one. I edited it in photoshop to get rid of the water marks (if this was for a game I would have simply traced over it or made my own). Then I made an alpha out of it to use in Zbrush.

This is the final Tombstone. Nothing too crazy but has some nice detail on it. I think I would have looked better if the extrusions were actual geometry and it wouldn't have bumped the poly count up much. As it stands, I have normal maps for the inward extrusions and normal maps for the flower pattern overlapping. Looks ok, but might change that when I put it all together.

Upate:  I actually went back and made the inward extrusions physical geo. looks much better. Only added 20 polys so totally worth it. I'll remember that for the next ones.

Here's another iteration using the same original base mesh for the cross.

This one has the same floral pattern, but in the centre it has a Celtic knot.

Update 02/06/16

I started working on the mausoleum last night, as I wanted a bit of a break from doing Tomb Stones.. I'll make some more once this is done. Couple more touches on this one then I'll unwrap it. After that I'll crate a high poly version on which I'll add some more detail in Maya before exporting it to Zbrush for some organic modelling (maybe a family emblem above the door etc)  

UV unwrapping this guy took longer than expected.. mainly due to the areas that I marked (the extrusions around the entrance and the top of the building).. I think it came out aright though in the end.

The previous method I used of importing my low poly mesh to Zbrush, then Zremeshing its on a high level before subdividing didn't do a great job here. It's still a method I will use for simpler geometry, and I'll probably try it with most models in the future as it's a great, efficient method. However, in this case it cave me some unwanted results, which I pretty much expected. Although some of the results could technically add to the warn look of the edges of the stone, I don't want to leave anything in this project that was unintentional. If I'm going to add such detail it will be planned. 

So to gain a little more control, I did some prep work in Maya first by adding support edge-loops to the hard edges of my model. 

Subdividing it in Maya gave me the results I wanted after adding the support edges, so I know it will work fine in Zbrush. I will be able to subdivide this mesh without the need of using Zremesher and my geo shouldn't deform. 

Time well spent. Took another 30 minutes or so, but was well worth it. 

Next problem was that even at a crazy 17million polys, the draw resolution still looked really low. I figured why though. When subdividing a mesh, it obviously keeps the same polygon flow but adds more geometry. So even though the poly count is through the roof, the polys arent square or evenly divided meaning in some place thousands of polys could be cramped into a small space, and an space of equal size could have less polys and be much less dens somewhere else on the mesh. 

So the solution was to Dynamesh the model I had prept for subdivision first. The prep work in adding the additional edge loops was still essential as the mesh would have otherwise completely deformed during the Dynamesh operation. However, if a surface is too thin, it will either be destroyed, leaving holes in the mesh, or will generate errors like in the picture above. 

As a quick fix, I swapped the walls with cubes instead of planes in Maya, ensuring the scales are 100% identical (to prevent errors with the bake later on)

Finally, after hours or trying to work out why things were going wrong, the draw resolution is fixed and so are the walls. 

So to recap - Prep the model the way you would before subdividing. Ensure the surfaces aren't to thin (could delete hidden faces and save the Low Ply version, then undo what you have deleted and leave the exta faces in on the High Poly version). Finally, Dynamesh the model to give a nice, even geometry before subdividing in Zbrush for some additional geo to work with. 

I added some detail to the Mausoleum, but I did this by using Alphas I'd made in photoshop. As I didn't do any sculpting, I could have just crated these alphas and pulled out the detail using Height in Substance Painter. These could have been applied directly to the Low Poly mesh. So no need to pre the model, bother with any export issues or deal with long Bake times from the high poly Mesh. Next time, I'll think longer about what it is I will be doing in Zbrush before Importing the model. If I had done some actual sculpting, this would have been the right move as, but for Alphas, stick to Substance Painter and only a low poly model. 

Update: 07/06/16

I ran a quick test today. I figured out that it would be much quicker to create my alphas in Zbrush (if I'm sculpting them like with the leaf) or Photoshop (if I'm creating them using existing images I've edited) and applying the detail directly to the low poly mesh in Substance Painter. This means I don't need to create a high poly mesh and bake the details every time. So again, if I'm  sculpting detail into a model, such a character or an organic object, then it makes sense to Sculpt a High poly mesh in Zbrush. If I'm only adding detail, even if it's cracks, wear or damage, I can use alphas in substance painter. 

Update: 28/06/16

Today I had some spate time and I thought I could quickly knock out the tree for the Graveyard. I had no concept art, although I was going for a stylised, spooky dead tree. The result was that I created a tree that I really wasn't happy with. I had a rough idea where I was going with it, but I decited to scrap this one and try to do the concept art myself. 

 I knew I wanted it to look old and twisted and have a swollen base on the trunk, almost like a pop belly. So I made a start by doing some different silhouettes. I made the first one, then copied it and altered it a little then did the same again until I was happy. This is the first time I have used this technique and it was a lot of fun as I was able to put to paper the mental image of the tree I had in my mind, and make some additional changes. Next I'll add some detail. I have decided that I am going to sculpt this tree in Zbrush as it will be the only tree in the scene and I want it to look awesome. 

Later I chose silhouette 4, then cut out pieces of reference images and used Warp and Puppet Warp in Photoshop to push things into place. This helped to give the tree the sort of twisted bark texture I wanted.

I then drew over some bits added a little more detail before adding some noise to fill out the tree. 

This is pretty much exactly what I had in my mind with some additional changes as I went a long. Using this method didn't take too long and I was able to create an image to give me some direction during the modelling process. 

Update: 29/06/16

I used the concept art I made and started modelling the tree today.
I started off by importing my Images to Zbrush and blocking out the shapes with Zspheres before adding an adaptive skin. I moved everything into place on a low subdivision and when I was happy with it I increased the polycount to several million and started sculpting. I now have the High Poly Mesh ready for retopology in Maya.

  

I already knew I was rarely happy with my work when it hadn't been based on concept art, but I have learned that I able to use tools in Photoshop to help me create my own concept art. Even if it's not to a high standard, It helps to give me direction and my finished work benefits from the extra effort. 

I tried to do the retopology in Zbrush, as I know how important poly flow is for animation etc, but I wanted to see if there was a quicker way of getting a low poly mesh out of a high poly environment model. 

After a lot of messing about and researching I found out that you can use poly paint on a Zremesh to determine the poly density of a certain area. The branches kept getting destroyed due to their harsh angles, so Increased the poly count in the problem areas and lowers it in others (Dark red is high, grey/blue is low)

I then used my new 900poly mesh, brought it into maya and made the high poly mesh the Live object, as if I were to quad draw. I then matched them both to the intended size for the game engines, hit quad draw and relaxed the polys on the low poly mesh around the live object. 

As I increased the poly count in the Zremesher to keep form, I deleted the extra edge loops around the branches ensuring I didn't create any N-gons or Tris. 

Considering I used a much more efficient way of retoplogising this mesh (which I would never use for characters) The results aren't too terrible. There is a tool in Zbrush to help guide your poly flow which I don't quite get... but once I nail that the topology should be much neater and the UV unwrapping won't be as much of a pain as it was.

Above is the low poly model with softened normals

This is the same low poly mesh after I baked the maps

QUICK PROGRESS CHECK

I uninstalled Excell and forgot, so had to scribble. 

Put the assets together in a scene to get a feel for how it's coming together.