README.md

---
title: Echo-Memory Action World Model
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.19.0
app_file: app.py
short_description: Action-conditioned video world model demo
python_version: "3.12"
startup_duration_timeout: 1h
---

# Echo-Memory: Action-Conditioned World Model Demo

Generate a video from an initial frame, a text prompt, and a camera rotation action using the [Echo-Memory](https://huggingface.co/papers/2606.09803) model.

**Model:** [Echo-Team/Echo-Memory](https://huggingface.co/Echo-Team/Echo-Memory) (Wan 2.1 1.3B backbone, context_k1 memory)
**Paper:** [Echo-Memory: A Controlled Study of Memory in Action World Models](https://arxiv.org/abs/2606.09803)
**Code:** [GitHub](https://github.com/Echo-Team-Joy-Future-Academy-JD/Echo-Memory)

## How it works

1. Upload an initial frame (or use one of the provided examples)
2. Enter a text prompt describing the scene
3. Choose a camera rotation direction and angle
4. Click "Generate Video" to produce a 5-second video clip at 640×352 resolution