More AI stuff.

[pwillard]

With the interesting discussions lately about making AI versions of train sim images, I got to thinking.

Does AI, in its current state, have the ability to help us with content? Maybe?

Here is a quick discussion:

Query: So if I have a scenery model of a building (movie theater) that is 20m x 18m x 9m with a mesh file name of "Citytheater.s" using an image file named "citytheater.ace", what would the SD file look like?

Great, that’s a very clean, concrete example 👍

Let’s build a correct, minimal, and scenery-appropriate Citytheater.sd for MSTS / Open Rails.

I’ll assume:

• Scenery object (not rolling stock)
• Meters (MSTS uses meters internally)
• Origin at ground center (typical for buildings)
• Dimensions: 20 m (W) × 18 m (D) × 9 m (H)

---

1️⃣ Bounding-box math (important)


For an ESD_Bounding_Box, you specify min and max corners.

Assuming:

• X = left/right (width)
• Y = vertical
• Z = front/back (depth)

Half-dimensions:

• Width: 20 ÷ 2 = 10.0
• Depth: 18 ÷ 2 = 9.0
• Height: 9.0 (grounded at Y=0)

So:

Min: -10.0 0.0 -9.0 Max: 10.0 9.0 9.0

---

2️⃣ Minimal but correct Citytheater.sd

SIMISA@@@@@@@@@@JINX0t1t______
Shape ( Class ( "Scenery" )
 Align ( None )
 Shadow ( 1 )
 Description ("City Movie Theater" )
 ESD_Bounding_Box ( -10.0 0.0 -9.0 10.0 9.0 9.0 ) )

✅ This is 100% valid for:

• MSTS
• Open Rails
• Shape Viewer
• TSRE

Hmmm... maybe there is hope for being truly lazy. (Though, we really don't need a bounding box on that... )