I wanted to break down my first AA quality model in a dedicated article, in order for me to track down my progress as a 3D artist, and for anyone reading this, hopefully finding it useful.
My dream was to become a 3D artist, preferably an environment artist. In order to achieve this goal, you should have a strong portfolio, communication skills, ability to receive a feedback (even the negative one), and passion for the art itself. Environments can range from small scenes to vast scenes, and for a beginner artist this may seem like an overkill task to perform. Thus, we approach the reason behind first building up portfolio that consists purely from props — it’s the way to demonstrate your ability to create high quality polished models, with as least difference or as much match to the reference image(s) as possible.
My very first attempt was the ancient british TV from 1950s, but I failed horribly at this task. The problem was my inability to do any of the industry standard workflows — high to low poly, texturing, subdivision modelling, etc. That way, I tried to make something simpler — I tried making classic GameBoy. Which as well turned out to be a failure. After 2 failures in a row, I felt kind of unmotivated and started to overthink it, and looking for reasons why to even bother learning 3D. Like they say,
Third time is the charm.
It seemed to be true for my case of failures in 3D modelling. For my third attempt of my first prop modelling I decided to pickup a simple butcher knife. Simple, yet complex, given the fact it’s my very first model of course. I can’t even measure or state how much I have learnt on this particular prop. I strongly recommend finding a group of people that are able to give you an honest, straight out of the heart critique of your work, with a good eye for details. For me, I’m grateful to be a part of such amazing community on a Discord. With that said, let’s have a look at my first prop that took me over 2 months to get finished.
I have to note that this timespan is very individual to everyone. In my case I was stuck at home during lockdown, being able to work seamlessly day after day on this prop and learn a ton.
The very first stop at the asset creation route is arguably the most important one, but you could say this about any part of this process. When deciding on what prop to create, I chose the one with enough reference images. Reason is absolutely simple — having enough reference images from all the different angles is essential for you to replicate all the details of the model and to keep the forms true.
There are several pages I used to look for references, but mostly it boils down to 3 big ones — simple Google Image search, Pinterest, PicClick. All of these provide you with mostly good references. Alternatively, the ultimate reference is to have a prop physically at home — that way you can take pictures from any kind of angle you need, and even look around it whenever you need to. This way you’d possibly save time on this phase of the project.
In case I didn’t find some angles I needed, I simply substituted it with similar-enough pictures to replicate desired part of the model.
As you can observe in the image within this section, I’m using PureRef which is a spectacular software for 3D artists or artists of any kind. It allows you to save all the images in one file and make additional edits on them as needed — rotate, scale, move, add notes, shuffle, and much more.
After gathering all the reference images, I was ready for another phase of the project — blockout. Blockout is the first step in creating the prop — it’s all about getting the dimensions right. It served for my future self to turn this blockout into high poly model, which is next step in the process. I’m a Maya user which means I used it for all the modelling part of this project.
In regards of the blockout phase and the amount of detail, I was trying to keep it not too detailed, instead only to catch the important dimensions of the knife and be ready for the high poly phase.
High poly phase means I’m trying to reach the highest amount of detail possible. With that in mind, by using subdivision modelling I started turning blockout into the high poly. Thinking of subdivision modelling as literally “melting your model,” and support loops to constraint the amount of melting, I considered using support loops on the edges where necessary in order to keep the constant offset of the edge. One particular part of the handle was an interesting case of variable bevel, which made me adjust the different support loops and distances from the edges. The result had over 500K faces.
Additionally, I applied different materials with distinct colors for different parts of the knife — blade, handle and rivets. This way I created ID map, using which the baker will be able to bake different parts of the knife on to the low poly.
Next step is to optimize the model for game ready use. That being said, I started deleting all the support loops or any edges that didn’t add up any difference to the silhouette of the final model. This phase of the model is meant to be the most optimized version of the model — as few polys, meshes and materials as possible due to drawback calls through the game runtime. The final result has around 600–700 polygons, single material and single mesh. I was very happy with such result, as the topology was optimized well enough. Final low poly mesh was triangulated, as I learnt different programs do triangulation differently, so it’s a lot safer to triangulate the mesh before proceeding any further.
Another important lesson I learned on this model is to keep the amount of long thin triangles as low as possible as this is a bad practice and results in worse performance.
Next step is to layout your UVs. In this phase of the project, I learnt a lot of new stuff as UVs were a huge mystery to me. For this project I used single 2048x2048 map, with 16 pixels of padding. The key is to keep the amount of your UV shells as low as possible, and their overall scale as huge as possible. That allowed me to use as much of the texture space as possible.
My final UV layout is following, which is not the best, I used only 79% of the UV space, but by far this was the best I could come up with. For this mesh I decided to cut it in half and mirror one side, easily cutting down required UV space by half. The mirrored UVs were then stacked on top of the original ones, and offset by 1 UV space to left.
With all my work done, I was happy to finally reach baking phase of the project. Baking is a process of getting the details of high poly model baked to the low poly mesh via the normal map. When I learned this I was literally blown away by how smart do people get with normal maps and it’s usage.
For this project I used Marmoset Toolbag 4 to bake, which is widely regarded among the best bakers out there. Baked mesh maps were afterwards imported into Substance Painter where I continued next stage of the process — texturing.
This may have been one of the longest phases of this project due to my lack of experience with texturing and knowledge of Substance Painter as a texturing package. This model however taught me how to do it better, and how to do it more properly. Big shoutout to Dennis Porter from 3DFT, who dedicated his time and thoroughly explained the process of texturing as a professional texture artist would tackle this prop. Texturing workflow should represent the creation of the prop in real life — when it’s first produced, assembled, what processes are used to refine it, what different surfaces occupy space on it, etc. Based on this knowledge I applied previous information and used it to texture the knife.
My mistake was using hand painting tools, instead of letting software do the job for me. The trick was using the masks, filters and generators to produce desired result. After battling with myself and Painter for a few weeks I finally reached the point where I was able to export the textures to Marmoset Toolbag 4 for the final renders.
Rendering & final post
After such a long time spent on this prop, I finally reached final stage — rendering the final screenshots and publishing it on my ArtStation page. The key in this stage is to setup a nice looking scene that will bring out all the detail on the prop. I setup a simple scene with blurred background and 3 point light setup and started playing with the camera settings, light settings and trying to bring it to life with the Toolbag.
Spending 3 or 4 months on this prop alone, I don’t regret a single hour. Honestly, I made a list of things and techniques I learnt just on this prop and the final number was 20. I also perfectly realize that there’s a lot more things to learn over time in order to improve as an artist but also as a person.
Special thanks to 3DFT community, which out of bottom of my heart — changed my life. Thanks to them, I learnt how to implement various different techniques in the 3D asset creation workflow. I also learnt the most important thing — to remain humble and be able to receive a critique without trying to defend myself or having any stupid excuses.