Week03

This week, my work mainly focused on exploring a wider range of visual possibilities and experimenting with the integration of AI tools into my material production workflow. Rather than aiming for final-level assets, the intention was to test different approaches and evaluate which methods are viable and effective within the context of the project.

At an early stage, I experimented with adding a “ground moss” pass to the sofa material, with the aim of visually blending the furniture more naturally with the surrounding vegetation used in the VFX elements. To support this idea, I also created a ground texture. However, several technical limitations emerged during this process. The sofa model has a high polygon count, which caused Maya to frequently freeze, and since the UVs were already finalised and aligned, further UV-based adjustments were not practical. I then attempted to extract positional data in Houdini to generate an alpha mask, but the resulting mask appeared irregular and unusable. As an alternative, I tested a triplanar projection workflow, directly projecting a noise-based mask onto the sofa. While this approach was technically functional, the visual result was not convincing and appeared more distracting than cohesive, so this direction has been temporarily set aside.

For the vegetation assets, I began exploring the use of AI in leaf texture creation. In general, plant assets—especially leaves—are highly dependent on the quality of texture sources, as texture fidelity plays a major role in perceived realism. Real photographic references often provide the most reliable results. I sourced high-resolution reference images of fern leaves and imported them into My GIF to generate PBR textures. The output quality was highly satisfactory, and for small-scale leaf assets, the results were more than sufficient. These textures were then imported into Substance Painter and placed using projection techniques, completing the leaf material efficiently.

Based on feedback from the previous week regarding overly uniform colour usage, I also revised the flower materials by introducing subtle purple variations. The petals were reworked on the final model with corrected UVs provided by a team member. Compared to earlier test versions, this iteration placed greater emphasis on fine details, such as petal veining and colour variation along the edges. As a result, the overall appearance differs noticeably from the initial tests, reducing the flat look and achieving a more realistic material response.

Next week, I will continue to develop and refine the flower materials, with a particular focus on exploring the range and presentation of emissive elements, as well as addressing the current issue of the petals appearing slightly grey and lacking vibrancy.