Medical Injector
A personal project focused on refining my hard-surface modeling, texturing, and real-time rendering skills through the creation of a near-future medical injector.This project explores a clean sci-fi design language, combining functional industrial forms with subtle storytelling through materials.
Special attention was given to surface wear, roughness variation, and balancing realism with stylized readability.
The workflow covers the full pipeline, including modeling, UVs, baking, texturing, and final rendering. A strong emphasis was placed on material definition and lighting to highlight form, edge detail, and surface quality.Thanks to everyone who shared feedback and helped push this project further. I hope you like it!
Modeled in Plasticity, low poly created in Maya, UVs in Maya, baking in Toolbag, texturing in Painter, rendering in Toolbag, staged with Megascans, and final compositing in Photoshop.
1 × 4K texture set: Albedo, AO/R/M, Normal
~21K tris
Workflow:
1. AI Concept Art Genertaion:
Different models show clear differences in how they handle creative interpretation versus visual correction.
ChatGPT is stronger in structural logic and creative reasoning.
Yuanbao is better at visual realism and material refinement based on real user photos.
Pure image-generation models have limited performance in engineering-oriented structural design scenarios.
The final workflow was:
Concept and logic development with ChatGPT → material and visual refinement with Yuanbao → manual engineering rationalization.
Below are the iteration process and results:
① Stage 1: ChatGPT merged a large number of reference images based on the prompt and continuously refined the output.
② Stage 2: A perspective view was selected, and Yuanbao was used to refine realism and material quality.
③ Stage 3: The orthographic views and product description were generated by ChatGPT, with manual adjustments made during modeling to correct any impractical or non-engineered details.
2. Modeling in Plasticity
Based on the AI-generated orthographic views, a high-poly model was created using Style3D and Plasticity.
3. Deres to Game-Ready Model in Maya
No AI tool has been found yet that can efficiently handle retopology for triangulated meshes, so Ngon cleanup and edge refinement were done manually in Maya.
MOI3D is recommended, as it can efficiently handle filleting and cleanup for models exported from Plasticity and other CAD-based modeling tools.
4. UV Unwrap in Maya and Rizom UV
The asset was unwrapped using a hybrid workflow between Maya and RizomUV. Initial seam definition and shell grouping were handled in Maya, followed by advanced packing and distortion optimization in RizomUV. Texel density was normalized across components, with attention to hard-surface edge continuity and minimizing stretching in high-visibility areas.
5. Baking in Marmoset
High-to-low poly baking was performed in Marmoset Toolbag, generating Normal, AO, and packed ORM maps. Cage adjustments and skew correction were applied to ensure clean projection, especially around tight mechanical details and chamfered edges. Meshes were split into logical bake groups to avoid intersection artifacts.
6. Texturing in Substance Painter
Texturing was carried out in Substance 3D Painter using a layered PBR workflow. Materials were constructed with a combination of smart masks and custom generators to simulate wear, edge damage, and surface variation. Special attention was given to material contrast between hard plastics, metals, and translucent fluid components to enhance readability.
7. Staging and Rendering in Marmoset
A studio-style setup and an outdoor HDRI were used in combination with key and rim lighting to enhance form readability and surface definition. Camera framing and post-processing—such as tone mapping, subtle bloom, and sharpening—were carefully refined to achieve a clean, product-visualization quality render.
