Edit model card

Paper

arxiv.org/abs/2405.20222

Introduction

This repo provides the inference Gradio demo for Hybrid (Trajectory + Landmark) Control of MOFA-Video.

Environment Setup

cd MOFA-Hybrid
conda create -n mofa python==3.10
conda activate mofa
pip install -r requirements.txt
pip install opencv-python-headless
pip install "git+https://github.com/facebookresearch/pytorch3d.git"

IMPORTANT: Gradio Version of 4.5.0 should be used since other versions may cause errors.

Checkpoints Download

  1. Download the checkpoint of CMP from here and put it into ./models/cmp/experiments/semiauto_annot/resnet50_vip+mpii_liteflow/checkpoints.

  2. Downloading the necessary pretrained checkpoints from huggingface. It is recommended to directly using git lfs to clone the huggingface repo. The checkpoints should be orgnized as ./ckpt_tree.md (they will be automatically organized if you use git lfs to clone the huggingface repo).

Run Gradio Demo

Using audio to animate the facial part

python run_gradio_audio_driven.py

Using refernce video to animate the facial part

python run_gradio_video_driven.py

IMPORTANT: Please refer to the instructions on the gradio interface during the inference process.

Downloads last month
0
Inference API
Unable to determine this model’s pipeline type. Check the docs .