Generate a scene with AI, then refine by hand

The AI Scene Generator extension takes a natural-language prompt, OpenSCENARIO XML, or DSL input and emits an editable canvas.

Natural-language scenario → editable scene on the canvas.

1. Open the AI Scene Generator

1. Click Extensions in the bottom-right of the canvas.
2. Pick AI Scene Generator.
3. Choose your AI provider (Anthropic Claude, OpenAI GPT, or Google Gemini) and paste your API key. The key stays in your browser.

2. Describe the scene

Paste a prompt like:

A 3-lane highway going left-to-right. An ego sedan (blue) in the centre lane, a truck (grey) in the right lane slightly ahead. Show a dashed path for the ego vehicle changing to the left lane.

Click Generate. The extension lays out lanes, vehicles, and a Path shape on the canvas.

3. Refine

Everything generated is a normal shape. Try:

• Selecting the truck and changing its template to Bus.
• Dragging the ego vehicle’s path to a different curve.
• Adding a pedestrian on the shoulder.

4. Or start from OpenSCENARIO

In the same panel, switch the input mode to OpenSCENARIO XML and paste a .xosc file. The extension lays out actors and trajectories on the canvas — the fastest way to visualise an existing scenario.

An ASAM OpenSCENARIO DSL example placed on the canvas.

What you learned

• AI generation is a starting point — every shape afterwards is editable like anything else on the canvas.
• The same panel accepts natural-language prompts and OpenSCENARIO input, so you can move between rough idea and formal spec without leaving the editor.

For prompt patterns that work well, see the AI Scene Generator README.