Seoul World Model

---
name: seoul-world-model
description: Skill for using the Seoul World Model — a world simulation model grounded in a real-world metropolis (Seoul) by Naver AI
triggers:
  - seoul world model
  - world simulation model
  - grounding world model in real city
  - street view world model
  - naver seoul simulation
  - urban world model inference
  - seoul street view generation
  - metropolis world model
---

# Seoul World Model

> Skill by [ara.so](https://ara.so) — Daily 2026 Skills collection.

## What Is Seoul World Model?

**Seoul World Model** (by Naver AI) is a research project that grounds world simulation models in real-world urban data from Seoul, South Korea. It enables:

- **World simulation**: Generate realistic video continuations of street-level scenes in Seoul
- **Street-view interpolation**: Synthesize smooth video transitions between street-view frames
- **Urban scene understanding**: Leverage a large-scale real-world metropolis dataset for training/evaluation

The project provides:
- Model checkpoints for world simulation inference
- Synthetic training data (Seoul street-view)
- Street-view interpolation model code and checkpoints

> ⚠️ **Note**: As of March 2026, the repository is undergoing internal review. Model checkpoints, inference code, and training data are planned for release. Monitor the [project page](https://seoul-world-model.github.io/#tldr) and repository for updates.

---

## Installation

### Clone the Repository

```bash
git clone https://github.com/naver-ai/seoul-world-model.git
cd seoul-world-model

Python Environment (Recommended)

# Create and activate a conda environment
conda create -n seoul-world-model python=3.10 -y
conda activate seoul-world-model

# Install dependencies (once requirements.txt is released)
pip install -r requirements.txt

Common Deep Learning Dependencies (Anticipated)

Based on the project type (video generation / world models), install:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
pip install diffusers transformers accelerate
pip install einops timm imageio[ffmpeg] opencv-python
pip install numpy pillow tqdm

Project Structure (Anticipated)

seoul-world-model/
├── README.md
├── checkpoints/          # Model weights (to be released)
├── data/                 # Synthetic training data (to be released)
├── inference/            # Inference scripts (to be released)
│   ├── world_model.py
│   └── interpolation.py
├── train/                # Training code (to be released)
├── configs/              # Model and training configs
└── utils/                # Utilities

Key Concepts