zelda-the-cat

DreamBooth model for the zzelda concept trained by Sanderbaduk on dataset of cats.

This is a Stable Diffusion model fine-tuned on pictures of my mum's cat "Zelda" with DreamBooth. It can be used by using the phrase 'zzelda cat' in a prompt.

This model was created as part of the DreamBooth Hackathon 🔥. Visit the organisation page for instructions on how to take part!

<table> <tr> <td>One of the images used to fine-tune on<br>"a photo of zzelda cat on a chair"</td> <td>One of the images generated by the model<br>"a photo of zzelda cat in space"</td> </tr> <tr> <td> <img src="http://i.imgur.com/zFOzQtf.jpg" style="max-height:400px"> </td> <td> <img src="http://i.imgur.com/12Nilhg.png" style="max-height:400px"> </td> </tr> </table>

Description

This is a Stable Diffusion model fine-tuned on images of my mum's cat Zelda for the animal theme.

To experiment a bit, I used a custom prompt for each image based on the file name. This works, but does not seem to have made much of a difference. The model was trained on CPU after encountering issues with CUDA, taking around 2 hours on 32 cores.

It works a lot better locally than in the widget, where it tends to take a few more tries to get the right cat.

Usage

from diffusers import StableDiffusionPipeline

pipeline = StableDiffusionPipeline.from_pretrained('Sanderbaduk/zelda-the-cat')
image = pipeline().images[0]
image