Worked on deterministic image generation for the Tencent/Hunyuan3D-2 repository by integrating a seed parameter into the t2i_worker component. This approach enabled reproducible image outputs, addressing variability and improving both reliability and debuggability for users and automated pipelines. The implementation focused on seed-controlled workflows, aligning with best practices for reproducibility in image generation tasks. Using Python and leveraging skills in image generation and reproducibility, the work emphasized clean, traceable code through a focused commit. These changes positioned the repository for broader adoption of seed-based rendering, supporting consistent outputs and streamlined testing in production environments without introducing new bugs.