Instructions to use ilkerzgi/krea-2-ember-pixel-dispersion-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use ilkerzgi/krea-2-ember-pixel-dispersion-lora with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("krea/Krea-2-Turbo", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("ilkerzgi/krea-2-ember-pixel-dispersion-lora") prompt = "a child flying a kite. ember pixel dispersion style" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Ember Pixel Dispersion
fal · Krea 2 Style LoRA series. One of 1600+ style LoRAs trained on fal.
A Krea 2 style LoRA. Add the trigger ember pixel dispersion style to the end of your prompt.

- Prompt
- a child flying a kite. ember pixel dispersion style

- Prompt
- a desert dune at sunset. ember pixel dispersion style

- Prompt
- a desert dune at sunset. ember pixel dispersion style

- Prompt
- a snowy mountain village. ember pixel dispersion style

- Prompt
- an old stone bridge over a river. ember pixel dispersion style

- Prompt
- a portrait of an old man. ember pixel dispersion style
Quick start
| Parameter | Value |
|---|---|
| Trigger | ember pixel dispersion style |
| Recommended LoRA scale | 1.0 to 1.25 |
| Base model | krea/Krea-2-Turbo |
| Trained on | fal-ai/krea-2-trainer |
Inference on fal
Python
import fal_client
result = fal_client.subscribe(
"fal-ai/krea-2/turbo/lora",
arguments={
"prompt": "a lighthouse on a rocky cliff. ember pixel dispersion style",
"loras": [{"path": "https://huggingface.co/ilkerzgi/krea-2-ember-pixel-dispersion-lora/resolve/main/ember-pixel-dispersion.safetensors", "scale": 1.0}],
"image_size": {"width": 1024, "height": 1280},
},
)
print(result["images"][0]["url"])
JavaScript
import { fal } from "@fal-ai/client";
const { data } = await fal.subscribe("fal-ai/krea-2/turbo/lora", {
input: {
prompt: "a lighthouse on a rocky cliff. ember pixel dispersion style",
loras: [{ path: "https://huggingface.co/ilkerzgi/krea-2-ember-pixel-dispersion-lora/resolve/main/ember-pixel-dispersion.safetensors", scale: 1.0 }],
image_size: { width: 1024, height: 1280 },
},
});
console.log(data.images[0].url);
Train your own on fal
Every LoRA in this series is trained on fal-ai/krea-2-trainer (100 steps, learning rate 3.5e-4). Train your own style from 4 to 10 reference images:
import fal_client
result = fal_client.subscribe(
"fal-ai/krea-2-trainer",
arguments={
"images_data_url": "https://.../your-dataset.zip",
"trigger_phrase": "ember pixel dispersion style",
"steps": 100,
"learning_rate": 0.00035,
},
)
print(result)
License
Krea 2 Community License. See the license.
ComfyUI
A ComfyUI-compatible copy of the weights is included in this repo: ember-pixel-dispersion-comfy.safetensors. It is the
exact same LoRA with the key prefix remapped (base_model.model. to diffusion_model.) so
ComfyUI loads it directly. Download that file into ComfyUI/models/loras/ and use it like
any other LoRA. The original ember-pixel-dispersion.safetensors is for fal / diffusers.
- Downloads last month
- -