search

fudan-generative-vision / hallo

Hallo: Hierarchical Audio-Driven Visual Synthesis for Portrait Image Animation
https://fudan-generative-vision.github.io/hallo/
MIT License
9.49k stars 1.3k forks source link

About the pretrained checkpoints of Reference Net #176

Closed cvipym closed 3 months ago

cvipym commented 3 months ago

In the paper, you said weights of the spatial cross-attention modules of ReferenceNet was optimized. But, why both Reference Net and runwayml/stable-diffusion-v1-5 have same checkpoints? They have equal sha256 results:

Reference Net checkpoints: https://huggingface.co/fudan-generative-ai/hallo/blob/main/stable-diffusion-v1-5/unet/diffusion_pytorch_model.safetensors

Original SD 1.5 checkpoints: https://huggingface.co/fudan-generative-ai/hallo/blob/main/stable-diffusion-v1-5/unet/diffusion_pytorch_model.safetensors

cvipym commented 3 months ago

Oh, I am really sorry. I totally misunderstood your words. I solved it.

cvipym commented 3 months ago

.