Natalie Portman Sdxl Flux

🚀 Natalie Portman (SDXL + FLUX)

This project is a collection of LoRA models related to Natalie Portman, trained on different bases and with various parameters, enabling text - to - image generation of the actress in diverse styles and scenes.

🚀 Quick Start

Prerequisites

• Ensure you have the necessary libraries installed, such as diffusers and torch.
• You need a machine with appropriate GPU support (CUDA) for better performance.

Usage Steps

1. Download the model weights in Safetensors format from the Files & versions tab.
2. Use the following code example to generate images:

from diffusers import AutoPipelineForText2Image
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to(device)
# Here you need to load the downloaded LoRA model, for example:
# pipeline.load_lora_weights("path/to/your/lora/model.safetensors")
prompt = "Your specific text prompt here"
image = pipeline(prompt).images[0]
image.save("output_image.png")

✨ Features

• Multiple Model Versions: There are several versions of the LoRA models (e.g., FLUX v.2, v6.0 2004, v5.0 Juggernaut X, etc.), each trained on different datasets and with different parameters, providing a wide range of options for generating images of Natalie Portman.
• Compatibility: The models are tested on various base models like FLUX.1 [dev], Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3, and AlbedoBase 2.0, ensuring flexibility in usage.
• Customizable Parameters: You can adjust parameters such as LoRA strength, CFG, Clipskip, and the number of steps according to your specific requirements.

📦 Installation

Install Required Libraries

pip install diffusers torch

Download the Model

Download the model weights in Safetensors format from the Files & versions tab.

💻 Usage Examples

Basic Usage

from diffusers import AutoPipelineForText2Image
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to(device)
# Load a specific LoRA model, for example, FLUX v.2
# pipeline.load_lora_weights("path/to/flux_v2_lora.safetensors")
prompt = "High res waist up portrait photo of Natalie Portman directly facing the camera"
image = pipeline(prompt).images[0]
image.save("basic_usage_image.png")

Advanced Usage

from diffusers import AutoPipelineForText2Image
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to(device)
# Load v6.0 2004 LoRA model
# pipeline.load_lora_weights("path/to/v6_2004_lora.safetensors")
positive_prompt = "{Artstyle, Character and scene description in usual FLUX fashion}, <lora:natportman_2003_juggerX_xl_1_wocap-natxprtmn-000149:1.0>"
negative_prompt = "ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child"
image = pipeline(positive_prompt, negative_prompt=negative_prompt, num_inference_steps = 20, guidance_scale = 6.0).images[0]
image.save("advanced_usage_image.png")

📚 Documentation

Model Description

FLUX v.2

• Training Data: Trained on FLUX.1 [dev] with 84 photos of Natalie Portman with GPT - 4 captions.
• Testing: Tested on FLUX 1.dev (full), FLUX fp8, and FLUX nf4.
• Parameters: Use around strength 0.8 - 1.2. Distilled CFG 3.5 and CFG 1.0 (without negative prompt). Clipskip 1.
• Usage Example:
  • Positive: {Artstyle, Character and scene description in usual FLUX fashion}, <lora:natportman_2004_local_164_v2-000005:1.0>

FLUX v.1 (Portman in 2004)

• Training Data: Trained on FLUX.1 [dev] with 100 photos of Natalie Portman in 2004 with GPT - 4 captions.
• Testing: Tested on FLUX 1.dev (full), FLUX fp8, and FLUX nf4.
• Parameters: Use around strength 0.8 - 1.2. Distilled CFG 3.5 and CFG 1.0 (without negative prompt). Clipskip 1.
• Usage Example:
  • Positive: {Artstyle, Character and scene description in usual FLUX fashion}, <lora:Natalie_Portman_2004_FLUX_epoch_16:1.1>

v6.0 2004

• Training Data: Trained on Juggernaut X with 230 photos of Natalie Portman from 2003 - 2004.
• Testing: Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3, and AlbedoBase 2.0.
• Parameters: Use around strength 1.0 - 1.1. CFG 5.0 - 7.0. Clipskip 1. 10 - 40 steps.
• Usage Example:
  • Positive: {Artstyle}, {Character and scene description}, elxolsn, <lora:natportman_2003_juggerX_xl_1_wocap-natxprtmn-000149:1.0>
  • Negative: ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child

v5.0 Juggernaut X

• Training Data: Trained for Juggernaut X with 270 photos of Natalie Portman.
• Testing: Best used with Juggernaut X, also works with Juggernaut v7 or RealismEngine 2.
• Usage Example:
  • Positive: {Artstyle}, {Character and scene description}, natxportman, <lora:natportman_gpt4_juggernautX_2_wocap-merger_21_65_83_04_02_04-natxportman:1.0>
  • Negative: ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child

v4.0 80mb

• Training Data: Trained on 137 photos of Natalie Portman with an improved Lora model which has a much reduced size while maintaining likeness.
• Parameters: Put the keyword "natxportman" at the beginning of prompts. Use Lora with strength around 1.1.

v3.1

• Training Data: Trained on a subset of the 200 - image dataset from v3.0, with better balance and captioning.
• Parameters: Lora strength between 0.8 - 1.2.

v3.0

• Training Data: Retrained on 200 images with an improved Lora training method and model.
• Parameters: Keep weight at 1.05 - 1.1.

v2.0

• Training Data: Lora trained on 270 images of Natalie Portman with SDXL 1.0 base.
• Testing: Tested on Juggernaut XL 3.0.

v1.0

• Training Data: LoRA trained on 45 images of Natalie Portman with SDXL 1.0 base.
• Parameters: The recommended strength for ComfyUi is 1.1.

Download Model

The model weights are available in Safetensors format. You can download them in the Files & versions tab.

Use it with the 🧨 diffusers library

from diffusers import AutoPipelineForText2Image
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to(device)
# You need to load the specific LoRA model you want to use here
# pipeline.load_lora_weights("path/to/your/lora/model.safetensors")
prompt = "Your text prompt"
image = pipeline(prompt).images[0]
image.save("output.png")

📄 License

This model is licensed under the [bespoke - lora - trained - license](https://multimodal.art/civitai - licenses?allowNoCredit=True&allowCommercialUse=Image&allowDerivatives=True&allowDifferentLicense=True).

Additional Information

Tags

• text - to - image
• stable - diffusion
• lora
• diffusers
• template:sd - lora
• migrated
• photorealistic
• sexy
• woman
• celebrity

Base Model

The base model is black - forest - labs/FLUX.1 - dev.

Widget Examples