Natalie Portman (SDXL + FLUX)
<Gallery />
Model description
<h2 id="please-donate-buzz-for-flux-lora-training-!-yjvq563os"><span style="color:rgb(250, 176, 5)">Please Donate Buzz for FLUX Lora Training !</span></h2><p></p><h3 id="flux-v.2-:-lomw9h4ha"><strong><u>FLUX v.2 :</u></strong></h3><p>Trained on FLUX.1 [dev] with 84 photos of<span style="color:rgb(250, 176, 5)"> </span><strong><span style="color:rgb(250, 176, 5)">Natalie Portman with GPT-4 captions</span></strong>. <strong><span style="color:rgb(34, 139, 230)">Tested on FLUX </span></strong><a target="_blank" rel="ugc" href="http://1.dev"><strong><span style="color:rgb(34, 139, 230)">1.dev</span></strong></a><strong><span style="color:rgb(34, 139, 230)"> (full) and FLUX fp8 and FLUX nf4 !</span></strong> Use around strength <strong>0.8-1.2</strong>. Distilled CFG 3.5 and CFG 1.0 (without negative prompt). Clipskip 1. Can be used for example as follows:</p><p></p><p><strong><span style="color:rgb(130, 201, 30)">Positive</span></strong><span style="color:rgb(130, 201, 30)"> : {Artstyle, Character and scene description in usual FLUX fashion}, <lora:natportman_2004_local_164_v2-000005:1.0></span></p><p></p><h3 id="flux-v.1-(portman-in-2004)-:-etmj0rpok"><strong><u>FLUX v.1 (Portman in 2004) :</u></strong></h3><p>Trained on FLUX.1 [dev] with 100 photos of<span style="color:rgb(250, 176, 5)"> </span><strong><span style="color:rgb(250, 176, 5)">Natalie Portman in 2004 with GPT-4 captions</span></strong>. <strong><span style="color:rgb(34, 139, 230)">Tested on FLUX </span></strong><a target="_blank" rel="ugc" href="http://1.dev"><strong><span style="color:rgb(34, 139, 230)">1.dev</span></strong></a><strong><span style="color:rgb(34, 139, 230)"> (full) and FLUX fp8 and FLUX nf4 !</span></strong> Use around strength <strong>0.8-1.2</strong>. Distilled CFG 3.5 and CFG 1.0 (without negative prompt). Clipskip 1. Can be used for example as follows:</p><p></p><p><strong><span style="color:rgb(130, 201, 30)">Positive</span></strong><span style="color:rgb(130, 201, 30)"> : {Artstyle, Character and scene description in usual FLUX fashion}, <lora:Natalie_Portman_2004_FLUX_epoch_16:1.1></span></p><p></p><h3 id="v6.0-2004:-axysmcpjd"><strong><u>v6.0 2004:</u></strong></h3><p>Trained on Juggernaut X with 230 photos of<span style="color:rgb(250, 176, 5)"> </span><strong><span style="color:rgb(250, 176, 5)">Natalie Portman from 2003-2004</span></strong>. <strong><span style="color:rgb(34, 139, 230)">Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3 and AlbedoBase 2.0!</span></strong> <strong><span style="color:rgb(250, 82, 82)">Use with keyword : "</span><em><span style="color:rgb(250, 82, 82)">ntxprtman</span></em><span style="color:rgb(250, 82, 82)">"</span> </strong>. Use around strength <strong>1.0-1.1</strong>. CFG 5.0-7.0. Clipskip 1. 10-40 steps. Can be used for example as follows:</p><p></p><p><strong><span style="color:rgb(130, 201, 30)">Positive</span></strong><span style="color:rgb(130, 201, 30)"> : {Artstyle}, {Character and scene description}, elxolsn, <lora:natportman_2003_juggerX_xl_1_wocap-natxprtmn-000149:1.0></span></p><p><strong><span style="color:rgb(230, 73, 128)">Negative</span></strong><span style="color:rgb(230, 73, 128)"> : ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child</span></p><p></p><h3 id="v5.0-juggernaut-x-:-o14faak6k"><strong><u>v5.0 Juggernaut X :</u></strong></h3><p>Trained for Juggernaut X with 270 photos of Natalie Portman. <strong><span style="color:rgb(250, 82, 82)">Use with Juggernaut X</span></strong>, Juggernaut v7 or RealismEngine 2. <strong>Works best with Juggernaut X</strong>.<span style="color:rgb(250, 82, 82)"> </span><strong><span style="color:rgb(250, 82, 82)">Use with keyword : "natxportman"</span></strong>. Can be used for example as follows:</p><p></p><p><span style="color:rgb(130, 201, 30)">Positive : {Artstyle}, {Character and scene description}, natxportman, <lora:natportman_gpt4_juggernautX_2_wocap-merger_21_65_83_04_02_04-natxportman:1.0></span></p><p></p><p><span style="color:rgb(190, 75, 219)">Negative : ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child</span></p><h3 id="-tog8fbyl1"></h3><h3 id="v4.0-80mb-:-nr2g8ntxj"><strong><u>v4.0 80mb :</u></strong></h3><p>Trained on 137 photos of <strong>Natalie Portman</strong> with an improved Lora model which has a much reduced size while maintaining likeness. <strong><span style="color:rgb(250, 82, 82)">Put the keyword : "natxportman" at the beginning of prompts!</span> </strong>Use Lora with strength <strong>around 1.1</strong>.</p><p></p><h3 id="v3.1:-8ewrekiit"><strong><u><span style="color:rgb(193, 194, 197)">v3.1:</span></u></strong></h3><p><span style="color:rgb(193, 194, 197)">Trained on a subset of the 200 images dataset from v3.0 but that is better balanced and has better captioning (more refined GPT-4 Vision prompt). Lora strength between </span><strong>0.8-1.2</strong><span style="color:rgb(193, 194, 197)">.</span></p><p></p><h3 id="v3.0:-frixneqbb"><strong><u>v3.0:</u></strong></h3><p><span style="color:rgb(193, 194, 197)">Retrained on 200 images but with an improved Lora training method and model to yield higher flexibility while preserving likeness and GPT-4 Vision captioning. Keep weight at </span><strong>1.05-1.1</strong><span style="color:rgb(193, 194, 197)">.</span></p><p></p><h3 id="v2.0:-cvb41l96z"><strong><u>v2.0:</u></strong></h3><p>Lora trained on 270 images of Natalie Portman with SDXL 1.0 base. The lora is also tested on <strong>Juggernaut XL 3.0</strong>. Most of the sample images were created with Juggernaut other than the "...2004..." generated ones.</p><p></p><h3 id="v1.0:-rn2d3oukc"><strong><u>v1.0:</u></strong></h3><p>LoRa trained on 45 images of Natalie Portman with SDXL 1.0 base, The recommended strength for <strong>ComfyUi </strong>is <strong>1.1</strong>.</p>
Download model
Weights for this model are available in Safetensors format.
Download them in the Files & versions tab.
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.bfloat16).to(device)
pipeline.load_lora_weights('Keltezaa/natalie-portman-sdxl-flux', weight_name='natportman_2004_local_164_v2-000005.safetensors')
image = pipeline('High quality passport photo of a woman wearing a suit and tie looking directly at the camera with her mouth closed and a neutral expression. She is also wearing a delicate gold chain and some understated diamond earrings. ').images[0]
For more details, including weighting, merging and fusing LoRAs, check the documentation on loading LoRAs in diffusers