Yabo Zhang

I have the same question !

The weights of StableDiffusion v1.5 has been passed by `--pretrained_model_path`. Thus, I think the path to safetensors could be None, but got weird results :)

@actionless Using SD v1.5 as base model still gets bad results :) I replace it with [Realistic Vision](https://civitai.com/models/4201/realistic-vision-v20).

Thanks for your attention! Could you provide more details of your implementation (e.g., generated results or some core codes), so that we can find out the reason together?

Thank you for your interest and appreciation of our work. The details of evaluation datasets have been sent to your email ([email protected]). We will update the details in revised arxiv.

Sorry for the delayed response. We provide the evaluation data below. [controlvideo-prompts.xlsx](https://github.com/user-attachments/files/18486829/controlvideo-prompts.xlsx) For video-text pairs used in quantitative evaluations, we choose 25 videos from DAVIS dataset and use ChatGPT to...

Thanks for your inspiration! Our ControlVideo requires motion sequence of all frames (e.g., depth maps) and text prompt to produce a video. In order to achieve first-frame generation, I think...

Thank you for your interest and appreciation of our work. The details of evaluation datasets have been sent to you by email. We will update the details in revised arxiv.

Thanks for your attention! Actually, you should download the whole repo (https://huggingface.co/runwayml/stable-diffusion-v1-5), and the downloaded directory is required ckpt.

The dir `checkpoints/stable-diffusion-v1-5/` is the downloaded repo, and cloning https://huggingface.co/runwayml/stable-diffusion-v1-5 into `checkpoints` would be fine.