Overview

This workflow allows you to generate a simulation-ready 3D asset starting from a text description. The system uses generative AI models to create reference images, turn those into a 3D mesh, and then apply textures before moving into physics, collision, and simulation validation. This path is ideal for quickly prototyping assets when no source files exist.

Step 1: Asset Name

The Asset Name is the label used to keep your project files organized. It becomes the base name for the generated images, mesh files, and downloadable assets. Users can freely type names with spaces (for example, “office chair”), and the system will automatically format them into filenames (for example, “office_chair”) to keep exports consistent and organized.

Step 2: Generation Prompt

The Generation Prompt is the most important input. This prompt is used twice in the pipeline:

• First, we use an image generation model to create initial reference images.
• Later, together with the selected image, we generate a 3D mesh.

Best Practices for Prompts

<aside> 💡

• “Red fire extinguisher with hose and nozzle, standing upright, realistic detail”

• “Ripe yellow banana” </aside>

• Be descriptive: specify object type, material, shape, and context (for example, “ripe yellow banana on a white background, isolated, product photo”).

• Avoid ambiguity: short one-word prompts (“banana”) may work, but multi-attribute prompts yield higher-quality meshes and textures.

• Consider simulator needs: prompts that describe physical characteristics (wooden, metallic, rubbery) help the pipeline assign more realistic physics properties later.

Gemini will generate three candidate images. Each preview can be refreshed if the results are unsatisfactory. Select the image that best matches your needs to continue. Once selected, the chosen image will populate the large preview window and unlock the Generate Mesh step.