Pieralongia / Seceda — builder’s story
84 seconds · English transcript of the on-screen story

I sat on this peak. I wanted a 3D model to show people what it was like.

Before Codex, I tried Claude. I asked it to describe the rocks first, then build an interactive scene.

Then I switched to Codex. I continued with editable Blender scenes, from a blockout to a photo-based model.

The data had a blind spot. A 2.5 m surface grid could anchor the terrain. It could not describe the cliff faces.

Two scenes. Two kinds of evidence. I preserved the measured grid and labeled the photo-assisted additions.

One experiment stayed out. The AI preview broke the landscape into thin fragments. No mesh was integrated.

I changed the shape, then the camera. Broken fins, pale stone, and a clearer gap brought the model closer to the photos.

Editable. Still unfinished. Measured terrain and interpreted rocks. I’m still improving the model in Codex.

Keep the limits visible. I can distinguish measured terrain from photo-based geometry. The model is still an interpretation.

A cleaner model is still the goal. I want to keep improving it so I can show people what being there was like.