import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Sula1723/gta", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]gta Dreambooth model trained by Sula1723 with TheLastBen's fast-DreamBooth notebook
This model was trained on the official art images of the GTA 5 game, which usually appear on the loading screens. These arts include images of people, animals, and various objects that are usually part of a general picture or background. Stable Diffusion v1-5 was used for training, also about 30 concept images and 1000 regularization images for style training. Due to the use of Google Colab resources, you do not need to have a GPU and you can freely experiment on your concepts. To select the optimal settings before training, I recommend using this guide written by Nitrosocke.
To reference the art style, use the token: gta
Test the concept via A1111 Colab fast-Colab-A1111
Sample pictures of this concept:
- Downloads last month
- 31