Skip to content

ojassurana/SceneMakerAI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

26 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SceneMakerAI

Agent Skill Codex Install with skills

Turn any 2D image into a world you can look around in.

Give Codex an image, describe how you want the world to feel, and get back an interactive 360 preview.

Start with any scene. Then ITERATE with plain-English edits: add UFOs, turn a lake pink, or change the mood.

No 3D modeling. No scene setup. Just image in, world out, preview live.

Install

Method 1: ask your agent to install it.

Install this skill:
https://github.com/ojassurana/SceneMakerAI

Method 2: use the skills CLI with npx.

npx skills add ojassurana/SceneMakerAI --skill scenemakerai

Examples

Desert Mesa Moraine Lake
Desert mesa input image Moraine Lake input image
Prompt
Make a 3D world for this image. Add 2 big ufos in the appropriate places.
Prompt
Make a 3D world for this image. Make it a pink lake instead.
SceneMakerAI desert world preview SceneMakerAI lake world preview

What It Does

  • Turns one image into an interactive 360-style world.
  • Lets you describe edits in plain language.
  • Supports follow-up iterations after the first preview.
  • Automatically hosts a local preview URL.
  • Refreshes the preview with cache-busted URLs so edits show up promptly.

What It Does Not Do

  • It does not create walkable 3D geometry.
  • It does not output cubemap faces.
  • It does not treat generated unseen areas as factual reconstruction.

Technical Output

SceneMakerAI generates one 2:1 spherical equirectangular environment-map image and previews it with Pannellum. It is still a flat image file, but it is authored to wrap inside a 360 viewer sphere, not to behave like a normal wide scenic banner.

The final image is saved under the agent's current working directory:

outputs/scenemakerai/<run-id>/panorama.png

Skill Layout

SKILL.md
agents/openai.yaml
references/prompting.md
scripts/prepare_pano_canvas.py
scripts/create_pannellum_viewer.py

SKILL.md is the main Codex skill file. The scripts handle deterministic canvas preparation and Pannellum viewer generation.

Basic Use

Ask Codex to use the skill with an image:

Use SceneMakerAI for this image. Add a few helium airships around the skyline.

The expected final response includes:

  • the saved panorama path
  • the local Pannellum preview URL

Preview Behavior

The viewer helper copies the panorama into the viewer with a content-hashed filename and returns a cache-busted URL, for example:

http://localhost:8000/?v=714c78f44b70

This avoids stale browser or Pannellum cache after edits.

About

Turn any 2D image into a 3D world, then iterate your 3D world

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages