A DreamBooth-LoRA for Krea 2, trained on Krea 2 RAW and shown on Krea 2 Turbo. The samples below were generated with this LoRA on Turbo (8 steps).
Use the token ashchild to invoke the concept.
"A cinematic wide shot of an ashchild wandering through a bioluminescent neon jungle, surrounded by floating translucent flora and iridescent mist."
"A moody, high-contrast portrait of an ashchild sitting in a dusty Victorian library, illuminated by a single shaft of golden sunlight piercing through heavy velvet curtains."
"An epic digital painting of an ashchild standing atop a jagged obsidian cliff overlooking a swirling cosmic nebula and shattered floating islands."
import torch
from diffusers import Krea2Pipeline
pipe = Krea2Pipeline.from_pretrained("krea/Krea-2-Turbo", torch_dtype=torch.bfloat16).to("cuda")
pipe.load_lora_weights("lloydchristmas1231/ashchild-30")
image = pipe("A cinematic wide shot of an ashchild wandering through a bioluminescent neon jungle, surrounded by floating translucent flora and iridescent mist.", num_inference_steps=8, guidance_scale=0.0).images[0]
image.save("output.png")
checkpoint-1000/optimizer.bin 195 MBcheckpoint-500/optimizer.bin 195 MBcheckpoint-1000/pytorch_lora_weights.safetensors 192 MBcheckpoint-500/pytorch_lora_weights.safetensors 192 MBpytorch_lora_weights.safetensors 192 MBcheckpoint-1000/scheduler.bin 1.4 kBcheckpoint-500/scheduler.bin 1.4 kBkrea/Krea-2-Raw