How to reproduce the result of pose2video?

[boanz]

Hello. Thank you for your great work. I can't reproduce the result when using the pose2video model. Can you provide any advice to help me resolve this issue?

command python scripts/inference/video2video.py --sd_model_name fantasticmix_v10 --unet_model_name musev_referencenet_pose --referencenet_model_name musev_referencenet --ip_adapter_model_name musev_referencenet_pose -test_data_path ./configs/tasks/example.yaml --vision_clip_extractor_class_name ImageClipVisionFeatureExtractor --vision_clip_model_path ./checkpoints/IP-Adapter/models/image_encoder --output_dir /projects/MuseV/output --n_batch 1 --controlnet_name dwpose_body_hand --which2video "video_middle" --target_datas duffy --fps 12 --time_size 96

configs/tasks/example.yaml - name: "duffy" prompt: "(best quality), ((masterpiece)), (highres), illustration, original, extremely detailed wallpaper" video_path: ./data/source_video/pose-for-Duffy-4.mp4 condition_images: ./data/images/duffy.png refer_image: ${.condition_images} ipadapter_image: ${.condition_images} height: 1280 width: 704 img_length_ratio: 1.0 video_is_middle: True # if true, means video_path is controlnet condition, not natural rgb video

output https://github.com/TMElyralab/MuseV/assets/34409364/149fec6e-62cc-4865-bd99-5ee5caadbf52