Instructions to use Hippotes/Krea-2-Experiments with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Hippotes/Krea-2-Experiments with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Hippotes/Krea-2-Experiments", torch_dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
File size: 766 Bytes
2746b81 50eeaaa 2746b81 6ff5b75 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 | ---
license: other
license_name: krea-2-community-license
license_link: https://huggingface.co/krea/Krea-2-Turbo/blob/main/LICENSE.pdf
base_model_relation: quantized
library_name: diffusers
tags:
- comfyui
- safetensors
base_model:
- krea/Krea-2-Turbo
- krea/Krea-2-Raw
---
Experiments with Krea 2 :)
- pre-release Turbo model produce the exact same outputs as the officially released but with a little something hidden inside to be found
- **Refiner Neutering**: cancel the actions of the txtfusion.refiner_blocks from the model. From my limited testing it has a similar effet to using a txtfusion.projector bypass lora or the Conditioning Multiply node but without the color saturation or overcooked textures. Load the patch as a lora, always use at strength 1. |