import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("oron1208/OOO_Anima", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("Ava2000/Choco_Chocolate_Shop")
prompt = "masterpiece, best quality, 4 fingers and a thumb,"
image = pipe(prompt).images[0]Choco (Chocolate Shop)
- Prompt
- masterpiece, best quality, 4 fingers and a thumb,
- Negative Prompt
- text, long body, lowres, cropped, lowres, jpeg artifacts, (signature, artist name, watermark, username, english text, signature:1.4), simple background, conjoined, out of frame, out of frame, censored, bar censor,
Model description
The style of Choco.
I made this lora to try and reproduce the style of Chocolate Shop.
Illustrious v3: somewhat the same as version 2, but it is smoother and I have the feeling it is slightly better. It still needs the artist tag (choco (chocolate shop),) for the best result. I think the dataset is not consistent enough, but fixing that would take way to much time, and using the artist tag is faster.
Backgrounds that are not well tagged might create artifacts. Define a background or use simple background together with (signature, watermark, username, artist name, english text, text, dated, signature:1.5), in the negatives for best results!
Illustrious v2: I made this lora because I noticed the old Illustrious lora trained in my early days was lacking in strength, These new lora works better, but for the best result, you best use the artist tag: choco (chocolate shop), It enhances your works greatly, but each on their own will give a lesser effect.
The dataset is trained with text and frames (but they are properly tagged). So be sure to use negative tags to counter this. If you don't know any, take a look at the demo images as a start.
Backgrounds that are not well tagged might create artifacts. Define a background or use simple background together with (signature, watermark, username, artist name, english text, text, dated, signature:1.5), in the negatives for best results!
Trigger words
You should use ch0c0 to trigger the image generation.
You should use choco \(chocolate shop\) to trigger the image generation.
Download model
Download them in the Files & versions tab.
- Downloads last month
- 14
Model tree for Ava2000/Choco_Chocolate_Shop
Base model
nvidia/Cosmos-Predict2-2B-Text2Image