For this project, my goal was to create two unique and interactive TikTok filters. The first was a scream mask that would appear when the user opened their mouth, while the second was a donut filter that would have had donuts falling down the screen and jiggling when the user raised their eyebrows.
However, when importing my Blender files into Effect House, I encountered a challenge with the triangle count being too high. Due to the nature of augmented reality and its limitations, this caused performance issues, preventing the filters from functioning as intended. I attempted several solutions, such as using the decimate modifier, compressing both the images and files, and lowering the render settings. Despite these efforts, the filters became distorted and compromised the shape of the object after each adjustment. Even using blender's strict hierarchy and changing the order in which I used and overlapped these potential fixes led to abysmal looking results.
Because there was no way that my projects were going to be able to work well with effect house, I decided to scale back my approach and create simpler, more optimized Blender files. I focused on reducing the complexity of the models to ensure they would meet the performance requirements of Effect House. This included smaller starting sizes, limited subdivisions and image projection rather than texture painting. Fortunately, this new approach worked, and I was able to create two functional, albeit ugly filters that were both visually engaging and lightweight enough for TikTok.
Through this experience, I learned valuable lessons in asset optimization and how to adapt my creative process to technical limitations. While this project was filled with challenges, I am pleased with the final outcome and the mass of knowledge gained along the way.
Comments