README.md

---
title: FloodDiffusion2-Live
emoji: 🚶
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.28.0
python_version: 3.10.13
app_file: app.py
pinned: false
license: mit
short_description: Live motion generation with prompt and path controls
---

# FloodDiffusion 2

[Project](https://alayalab.github.io/FloodDiffusion2/) · [Code](https://github.com/AlayaLab/FloodDiffusion2)

Choose **SEED** or **HumanML3D**. Edit the four prompts and speeds while stopped, then **Start**.

Use **WASD**, the on-screen joystick or a gamepad's left stick to move. Press **1–4** to switch prompts. Drag or use the right stick to rotate the camera; scroll to zoom.

Target speed accelerates and decelerates smoothly. The ground path follows the target with spring damping; segment colors identify each prompt. Each Start begins a new motion with 30 history frames and a 30-frame active window. Each session runs for up to 55 seconds.

SEED shows a body mesh and skeleton; HumanML3D shows a skeleton.

Body visualization uses SMPL-Body, courtesy of the Max Planck Institute for Intelligent Systems, under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). See [asset attribution](space/assets/BODY_LICENSE.md).