Datasets for ControlNet Ghibli-Style Training? 🎨

🤗Datasets
1

Hi Hugging Face community! :waving_hand: I’m looking to train a ControlNet model for generating Studio Ghibli-style images.

Are there any datasets on Hugging Face or elsewhere optimized for this aesthetic, with ground truth images, conditioning images (e.g., Canny edge, depth maps), and prompts? :deciduous_tree:

Specifically, I need datasets that capture Ghibli’s vibrant, hand-drawn look. Any recommendations for datasets, preprocessing tips, or relevant Hugging Face Spaces? Thanks! :raising_hands:

1 Like
2

Isn’t normal LoRA sufficient?

1 Like
3

using ghibli controlnet for my project to show skill. :kissing_face:

1 Like
4

I see. I think many people create training datasets for ControlNet. For example, for Depth ControlNet training, you could use images generated by Depth Anything. I think you could create data for Canny without using an AI model, just with image processing.

1 Like