README.md

metadata

title: MotionLCM - Real-Time Text-to-Motion
emoji: 🏃
colorFrom: blue
colorTo: yellow
sdk: gradio
sdk_version: 6.10.0
app_file: app.py
pinned: false
license: mit
python_version: '3.12'

MotionLCM — Real-Time Text-to-Motion

Generate realistic 3D human motion from text in ~30ms.

• MotionLCM: 1-4 denoising steps, 1929x faster than baseline
• MLD Baseline: 50 DDIM steps, higher quality reference
• 22-joint skeleton from HumanML3D dataset

Author: Soumyanil Ain | MS CS | UNC Charlotte Paper: arxiv.org/abs/2404.19759 (ECCV 2024)