Tencent / MimicMotion

High-Quality Human Motion Video Generation with Confidence-aware Pose Guidance
https://tencent.github.io/MimicMotion/
Other
1.93k stars 165 forks source link

我有一组DWpsoe的图片,如何直接统统Dwpose生成视频呢? #51

Open DatanIMU opened 4 months ago

DatanIMU commented 4 months ago

求命令,用dwpose直接生成视频。

nitinmukesh commented 4 months ago

Create multiple YAMl files in config folder and change name of ref_image_path in each of them

Suppose youe created 3: 1.yaml, 2.yaml, and 3.yaml

Create batch file on Windows or sh on Linux and add

python inference.py --inference_config configs/1.yaml python inference.py --inference_config configs/2.yaml python inference.py --inference_config configs/3.yaml

This will process in queue. Make sure to activate virtual environment in command prompt and then launch batch/shell file

DatanIMU commented 4 months ago

How to write YAMl files? replace ref_video_path?

base svd model path

base_model_path: stabilityai/stable-video-diffusion-img2vid-xt-1-1

checkpoint path

ckpt_path: models/MimicMotion_1-1.pth

test_case:

haihaict commented 4 months ago

check comfyui implementation and its sample code, it would be super easy.

nitinmukesh commented 4 months ago

@DatanIMU

You need to update these. If you want to use the same video but different images, in that case only change image path. ref_video_path: assets/example_data/videos/pose1.mp4 ref_image_path: assets/example_data/images/demo1.jpg