Creating Models for Different Purposes with Level of Detail Generation
Ni haody ya’ll! I’m Hakim, a 3D Game Artist at Mighty Bear Games!
Working on Mighty Action Heroes posed some new challenges in the way we approached character creation. We had to make a few different sets of character models for different purposes, otherwise known as Level of Detail (LOD):
- A batch of high-quality models, ready to be used for marketing materials
- A set of character models to be used in the lobby screen
- A final variation of in-game character models that are well-optimized
If you check out our Marketing Artist Hazel’s article, you can see for high-quality models that are used for marketing purposes, extra detail and fidelity are needed. While models used in the main menu and in-game are identical, with the former’s version having additional details like fingers.
We needed to figure out the best solution for creating all three different assets in a smooth and efficient manner. The way in which we came up with the Character LOD generation pipeline was a collaborative effort, comprising cross-discipline knowledge sharing, numerous discussions, and camaraderie.
Challenges in Character Modeling
Early on, our character creation process began with brainstorming sessions where we defined what we needed for each Character LOD. Did we need additional text maps other than the base color? Did we need fingers for the in-game characters? What else can be reduced? This phase involved heavy collaboration between 3D artists, tech artists, and our animator.
We initially modeled the high poly characters in-house, to set down the technical requirements for them. With the technical requirements down, we recruited outsourcers to help produce the high polycount models. Check out more in our other 3D Artist, Irsyad’s article here! It's a good read about the considerations for asset creation.
From there, we had to take those high poly models and reduce them to be implemented in-game. When we first started this process, there was a lot of trial and error. We sought out tools that would enable the mesh to be automatically reduced. This did not prove to be effective as the models had a lot of issues, such as unclean topologies and broken silhouettes, on top of generally being very messy.
Since all our characters have the same starting base shape, we instead opted to model a new, base, low poly version of the characters. We started with the base character and from here, we would take elements from the high poly models and reduce them to match the poly count of the LOD0 base model. These would be parts like hair, shoes, accessories, etc. Basically, the stuff that makes up the character’s uniqueness.
Challenges in Texturing
Having the hands, head and other body parts as separate meshes helped with the creation process. However, creating the low poly versions was only half the battle won. We still had to figure out a way to transfer our texture maps over efficiently. For this, we utilized Maya’s in-built transfer maps tool. This takes the textures from the high poly model and reprojects them onto the low poly model’s new UVs. However, this method caused texture bleeding, where parts of the textures spill over into other areas. Necklaces will bleed into the necks, joint areas will be overly dark, etc.
Transfer maps provide better results, i.e. less bleeding if we can transfer all the different parts of the models separately. However, this was a time-consuming process as it meant we had to deal with many different texture maps that were created. After transferring the maps, we would need to bring both them and the model into Substance Painter and apply paint-over fixes until all the bleeding issues were fixed.
Next up, working on animation compatibility. Collaboration between the 3D artists and animators was essential to make sure every character could come to life seamlessly. Our animators were integral to this phase. They worked closely with the character modeling team to ensure that the characters’ designs were animation-friendly.
For this, we discussed the best way to test poses to create the files required with our Animator, Quinnie. We used an animation file with many different poses, to ensure that all our characters had paint weight issues and had no problems being used in any pose we wanted. Here’s what she had to say about optimizing efficiency in game art with AI-assisted art tools!
Tools and Collaboration
Once we had figured out the general workflow, we began listing down any pain points we had; areas that were repetitive, or things that were taking a long time. Anything that we wished could be solved with a few simple clicks, we noted down and discussed with our Technical Artist, Wayne Yip the GOAT!
- Transferring texture maps that were taking way too long
- Weight painting the model from scratch that was gonna take wayyy too long
- Exporting things with the correct hierarchy that was gonna take wayyyy too long
I’m exaggerating a little here but, you get the idea. One of our biggest challenges during this process was minimizing repetitive tasks. Our technical artist thus worked closely with us to craft specific tools we needed to help expedite our work. We had a tool to rename the different parts of the model, one to automate the transfer of maps, another to copy skin weights, and finally, one to combine and export completed models.
Testing and iterating these tools was a back-and-forth, open process between the 3D and tech art teams. Everyone was hands-on, helping to test and make things better. This was honestly a very fun stage as we got to test new tools and see how much faster our work was progressing.
Through vigorous testing and feedback, we were able to overcome obstacles and smooth out our workflow. We made sure to block out an appropriate amount of time for research and development. This helped prevent us from taking too much time to work on or troubleshoot a certain feature. We did most of what we could at any given time.
In conclusion, the journey of developing and fine-tuning our workflow, creating efficient tools, and ensuring animation compatibility was a testament to the power of human creativity and collaboration. The most remarkable and usable outcomes are achieved when talented individuals work together to push the boundaries of what’s possible. By identifying our challenges, devising effective workflows, and addressing pain points, we achieved the results we were looking for.
As we move forward, we remain vigilant in exploring new tools, whether AI-driven or not, to further enhance our workflow and elevate our creative process.
Creating Models for Different Purposes with Level of Detail Generation was originally published in Mighty Bear Games on Medium, where people are continuing the conversation by highlighting and responding to this story.