---
license: other
license_name: bespoke-lora-trained-license
license_link: https://multimodal.art/civitai-licenses?allowNoCredit=True&allowCommercialUse=Image&allowDerivatives=True&allowDifferentLicense=True
tags:
- text-to-image
- stable-diffusion
- lora
- diffusers
- template:sd-lora
- migrated
- photorealistic
- female
- woman
- celebrity

base_model: black-forest-labs/FLUX.1-dev
instance_prompt: 3mm4w
widget:
- text: '3mm4w,  photo of a woman''s face '
  
  output:
    url: >-
      24659741.jpeg
- text: '3mm4w, candid photo of a young woman wearing a yellow sports jacket and a graphic tee smiling and looking at the viewer, upper body focus '
  
  output:
    url: >-
      24659522.jpeg
- text: '3mm4w, candid photo of a young woman with blonde hair wearing a white turtleneck sweater looking at the viewer, pensive expression '
  
  output:
    url: >-
      24659715.jpeg
- text: '3mm4w, photo of a woman''s face'
  
  output:
    url: >-
      24423773.jpeg
- text: '3mm4w, photo of a woman wearing hogwarts school uniform holding a book looking at the viewer standing in the yard of medieval castle '
  
  output:
    url: >-
      24423776.jpeg
- text: '3mm4w, candid photo of a young woman with blonde hair wearing a white turtleneck sweater looking at the viewer, pensive expression'
  
  output:
    url: >-
      24424281.jpeg
- text: '3mm4w, photo of a woman wearing a red sweater siting next to a window overlooking a forest, sunlight spreads across her face as she smiles, cozy interior'
  
  output:
    url: >-
      24423774.jpeg
- text: '3mm4w, photo of a woman wearing a white shirt, leather jacket and jeans posing for a photograph in a garage'
  
  output:
    url: >-
      24423775.jpeg
- text: '3mm4w, photo of a young woman dressed in an elaborate Victorian-era gown with intricate lace and velvet detailing, is attending a grand masquerade ball. Her mask is adorned with feathers and gemstones, and her expression is a mix of intrigue and mystery as she glides gracefully across the ballroom floor. The setting is opulent, with chandeliers casting a warm glow over elegantly dressed guests. Illustrate this in a classical, romantic style with soft lighting and rich textures'
  
  output:
    url: >-
      24423777.jpeg
- text: '3mm4w, an early 2000s-style photo of a young glamorous woman posing on the red carpet at a movie premiere. She’s wearing a sleek, metallic silver evening gown , paired with strappy stilettos and oversized hoop earrings. Her hair is styled in voluminous waves with a side part, and her makeup features a smoky eye and glossy lips. The background is a bustling red carpet event with photographers snapping pictures, velvet ropes, and movie posters. The overall tone should reflect the vibrant, glitzy aesthetics of early 2000s Hollywood with warm, high-contrast lighting and a slight lens flare for that classic premiere effect.'
  
  output:
    url: >-
      24423779.jpeg
- text: '3mm4w, candid photo of a young woman wearing a yellow sports jacket and a graphic tee smiling and looking at the viewer, upper body focus'
  
  output:
    url: >-
      24423990.jpeg

---

# Emma Watson / Hermione Granger - [Flux, Pony, SD1.5] 

<Gallery />



([CivitAI](https://civitai.com/models/))

## Model description

<p><strong><span style="color:rgb(250, 176, 5)">IMPORTANT:</span></strong><br />The default model that's available for download is overtrained and will steer heavily toward the default appearance. The prompt adherence to words that change that is very low. To get a more versatile checkpoint, please use one of the other checkpoints that are included in the attached zip file - the lower you go, the more malleable it will get (likeness will also reduce by a bit).<br />Will rework the training workflow and try to get a better model out soon.<br /><br /><strong><span style="color:rgb(250, 176, 5)">Flux Retrained</span></strong><br /><span style="color:rgb(193, 194, 197)">Retrained model with the same dataset. I have uploaded the 4000th epoch but some of the earlier epochs might be better. Generating a single image takes &gt;2 minutes on my device so haven't been able to do a good comparison. </span><br /><span style="color:rgb(193, 194, 197)">Depending on the feedback on this version, will test and release earlier step.</span><br /><br /><strong><span style="color:rgb(250, 82, 82)">NOTE: I have added multiple checkpoints in a zip file to the model. The default version will have very strong likeness but if you'd like to test earlier steps for better variance, feel free to find a step you like. If you manage to get better gens out of any of the other checkpoints, please post them here and share your params in the comments.</span></strong><br /><br /><strong><span style="color:rgb(250, 176, 5)">Flux</span></strong><br /><span style="color:rgb(193, 194, 197)">Recommended strength is 1.</span><br /><span style="color:rgb(193, 194, 197)">This version was trained (21, 25 images) on Flux.1 Dev. Picked the last model at 4000 steps. Trained on an A40 on runpod - took 3 hours.</span><br /><br /><span style="color:rgb(193, 194, 197)">I tried running it on schnell but it didn't work, if somebody can figure out how to run it on schnell/schnell-dev-merge, that'd be great - reach out.</span><br /><br /><strong><span style="color:rgb(250, 176, 5)">Pony</span></strong><br /><br /><span style="color:rgb(193, 194, 197)">This version was trained (18, 23 images) - (24, 16 repeats) on Pony Realism. Picked the 6th epoch.</span><br /><span style="color:rgb(193, 194, 197)">The face likeness is pretty good IMO, further training did not improve it.</span><br /><span style="color:rgb(193, 194, 197)">I have included an image where I used facedetailer/inpainting for the face with my SD 1.5 model. This provides the best output IMO.</span><br /><span style="color:rgb(193, 194, 197)">Also added a comparison of Pony facedetailer vs 1.5.</span><br /><br />V2<br />LORA trained on 30 publicly available images of Emma Watson.<br /><span style="color:rgb(250, 176, 5)">This model comes with an optional trigger word </span><code>h3rm10n3</code><span style="color:rgb(250, 176, 5)">to replicate Hermione Granger from the Harry Potter movies.</span></p><p>Trained on base SD 1.5, should work with most models.</p><p><span style="color:rgb(193, 194, 197)">Use ADetailer if details are lost and strength of 0.8 should work on most prompts.</span><br /><span style="color:rgb(193, 194, 197)">Consider leaving a review if you like the model. All feedback is appreciated.</span><br /><br />V1<br />LORA trained on 90 publicly available images of Emma Watson.</p><p>Trained on base SD 1.5, should work with most realistic models.</p><p>Use ADetailer if details are lost and strength of 1 should work on most prompts.</p>

## Trigger words
You should use `3mm4w` to trigger the image generation.
    

## Download model

Weights for this model are available in Safetensors format.

[Download](/Keltezaa/emma-watson-hermione-granger-flux-pony-sd1-5/tree/main) them in the Files & versions tab.

## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)

```py
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/emma-watson-hermione-granger-flux-pony-sd1-5', weight_name='3mm4w-flux.safetensors')
image = pipeline('3mm4w, candid photo of a young woman wearing a yellow sports jacket and a graphic tee smiling and looking at the viewer, upper body focus').images[0]
```

For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)