G-Ray Space: 3D Generative Ai Web App
This image was generated with Nano Banana via Scenario
High poly results
High poly results wire
G-Ray Space comes with a Blender bridge plugin developed with Claude Ai. Rather than fully relying on generative ai models for retopology, I find it best at the moment to combine the automated steps with manual work in Blender.
This was the retopologized model with Quad Remesher by Exoside in Blender. Quad Remesher is a purchased add-on with license. Not included with Blender.
The retopologized model was then sent back to the web app and then textured with Tripo Texture 3.0 model.
G-Ray Space also comes with HDRI scenes and parameters allowing users to visualize the 3d model with standard capabilities.
These are the populated and functioning(Not perfectly) tabs. I set these up to follow the standard pipeline in 3d development with a mix of hands on approaches between the app and DCCs such as Blender.
This is the Helper, which handles the api keys locally in the users OS credentials manager. Currently focuses on windows.
wip documentation of the app, analytics, and console popups.
G-Ray Space: 3D Generative Ai Web App
G-Ray Space is an experimental web application I am developing utilizing AI in Visual Studio Code, primarily leveraging Claude for implementation. While AI handles much of the code generation, the direction, structure, and functionality of the app are guided by my experience as a 3D artist. My understanding of production pipelines, DCC tools, and real-world workflows shapes how the system is designed and evolves.
The project began as a single HTML file containing all styling and scripting. As the application grew, I transitioned it into a modular structure, separating components into organized files and systems that are dynamically loaded into the main interface. This shift reflects a move toward a more scalable and maintainable architecture.
G-Ray Space integrates generative AI models to create 3D assets and textures directly within the browser. While generation is handled in-app, I intentionally keep optimization and cleanup within traditional DCC workflows, particularly Blender. To support this, I developed a custom Blender add-on (via Python/bpy) that acts as a bridge between the web app and Blender, allowing for a more seamless transition between AI generation and professional asset refinement.
A key feature of the platform is the built-in paint editor. While currently lightweight, it enables users to perform essential image editing and layering. These layers are consolidated into a single reference image when transitioning into the 3D generation workspace. This creates a more intuitive, artist-driven workflow that emphasizes visual iteration over more technical node-based approaches.
Security and data ownership are core priorities. G-Ray Space follows a BYOK (Bring Your Own Key) model, where all API keys are stored locally on the user’s machine via the operating system’s credential manager. A companion local server application, G-Ray Helper, handles API communication, key management, and real-time console feedback for generation processes. This ensures that sensitive data, including API keys, never leaves the user’s environment.
Currently, Scenario is used as a default provider to access multiple generative models, though users can configure their own API keys for other supported services. The platform itself acts purely as an interface and workflow tool, rather than a centralized service.