FMP – Simulation
For the final scene’s animation, I initially used Mixamo animations, but I felt they weren’t sufficient. So, I decided to use motion capture data instead. Mahadev helped me since I couldn’t use the motion capture app (Move AI, which is only available on iPhone). To be honest, I didn’t expect the quality to be very good, but when I checked the data in C4D, it worked surprisingly well, aside from a bit of flickering. There were some minor errors when I applied the mocap data to the Mixamo model, but I was able to resolve them through trial and error.
When using Mixamo animations, the workflow is straightforward: Mixamo animations already have keyframes on their skeleton, and you can create motion clips from these and blend them in the motion clip editor. However, with mocap data, the process is more complex. I needed to transfer the animation from the mocap skeleton to Mixamo’s T-pose skeleton using the character definition tag. In this case, the keyframes were not transferred to Mixamo’s skeleton. When I created a motion clip from the transferred data, no animation appeared in the clip. I searched for solutions but couldn’t find anything useful. Eventually, I asked ChatGPT. Although its response wasn’t entirely accurate, one suggestion helped: after baking the transferred animation onto the skeleton, I was able to create a motion clip and apply it to the original model.
During this process, I had a feedback session with Manos. I wasn’t sure what kind of animation to apply to the character, so I asked for suggestions. He proposed having the character morph into text. At first, I planned to show the character and text together, so I didn’t immediately follow his suggestion. However, after testing simulations with both the mocap animation and text, I realized the animation needed to be too long, and it didn’t feel meaningful. I decided instead to have the background pictures morph into the character, and then have the character morph into the text.
To achieve this effect, I explored different approaches to creating particle-based visuals in C4D. There are three main methods: cloner, matrix, and particle systems. Since the particle system feature is relatively new in C4D and lacks comprehensive tutorials, I couldn’t get it to work as intended. I switched to using matrix simulations, which function similarly to particle systems by creating particle instances based on objects. To morph one shape into another, I used the inheritance effector. However, since I used multiple effectors in the same project, they sometimes interfered with each other in unintended ways. Each effector needed to be applied to its specific Mograph object, but as it was my first time using this method, I occasionally forgot to apply them, causing the simulation to fail. After extensive trial and error, I finally managed to create the effect I envisioned.
With the draft outcome ready, I ran a test render. Unfortunately, several issues arose again during the rendering process. The first problem involved using matrix (cube) particles on the surface of the human model alongside a wireframe-textured human model. Since the simulation was based on the model, they were supposed to move identically, but in the rendered sequence, the simulation lagged by one or two frames. To resolve this, I baked the matrix simulation, as I had done for the emoji simulation, but the problem persisted. While adjusting the render settings, I discovered the frame offset for baked simulations. Setting the offset to -1 frame aligned the simulation with the model in the render view.
Additional issues occurred with the baking process itself. While the simulation worked fine in the viewport, exporting it as a Mograph cache sometimes caused errors. For example, certain simulations would unexpectedly use the base mesh from a previous animation. These issues only became apparent after rendering, which took about eight hours, resulting in wasted time. This problem happened multiple times. The solution was simple—clear the cache data and bake again—but it was time-consuming and frustrating, as each render and baking process took considerable time.
Now that the animations and simulations are adjusted, I’m preparing to render the final shot and composite everything together. I really hope there won’t be many more issues from this point onward!
Leave a Reply