Detailed Documentation

1. Setting Up Local AI with Ollama

Ollama allows you to run large language models locally on your machine. This ensures your 3D assets remain private and you don't incur any API costs.

• Installation: Go to ollama.com/download and install the appropriate version for your OS.
• Starting the Server: Open your terminal or command prompt and run ollama serve. Leave this terminal window open while using Blender.
• Downloading a Vision Model: Open a new terminal tab and run ollama pull llava:13b. LLaVA is a powerful open-source vision model capable of understanding images. This download may take some time depending on your internet connection.
• Blender Setup: In Blender, navigate to Edit > Preferences > Add-ons > AI Object Namer. Select "Ollama (Local)" from the AI Provider dropdown.

2. Setting Up Cloud APIs

If you prefer using high-end cloud models, you will need to provide an API key. Please note that using these services requires an internet connection and you may be billed by the provider based on usage.

Single vs Batch Renaming

The addon offers two primary modes of operation:

• Name Selected Object: Processes only the Active object in your selection. You can also use the Preview Name button to see the AI's suggestion before applying it.
• Name All Selected: A batch operation that loops through every selected object in the viewport, capturing and renaming them sequentially. Progress is usually printed to the system console.

Multi-Angle Capture Settings

Complex geometry can be difficult to interpret from a single isometric view. The addon can stitch together multiple views of your object into a single image strip before sending it to the AI.

Using Custom Templates

If you select the "Custom" naming style, you can define exactly how you want the final name formatted using placeholders. Available placeholders include:

• {name} - The core descriptive name generated by the AI.
• {type} - Often derived as 'MESH', 'CURVE', 'LIGHT', etc.
• {material} - Extracted from the object's active materials.

Example pattern: ENV_{name}_01 outputs ENV_Wooden_Barrel_01

AI Input Metadata Context

The AI can make more educated guesses if it knows more about the object. In the Context sub-panel, you can pass:

• Dimensions: Real-world bounding box scale (e.g. 2m x 1m x 0.5m) helps differentiate a toy car from a real car.
• Collections: Providing the name of the parent collection (e.g. "Kitchen_Props") sets strong scene context.
• Materials: Passing material names (e.g. "Mat_RustyMetal") helps the AI determine composition.

Custom System Prompts

For ultimate control, you can override the AI's internal instruction prompt entirely in the Advanced Capture Settings panel. This allows you to completely dictate the AI's persona, strict JSON outputs, or behavioral nuances.