---
base_model: black-forest-labs/FLUX.1-dev
widget:
  - text: yearbook portrait of student, smiling, crooked teeth
    output:
      url: images/8-1.png
  - text: yearbook portrait of student, smiling, crooked teeth
    output:
      url: images/8-2.png
  - text: yearbook portrait of student, smiling, crooked teeth
    output:
      url: images/8-3.png
  - text: yearbook portrait of student, smiling, crooked teeth
    output:
      url: images/8-4.png
instance_prompt: crooked teeth
---
# Crooked-Teeth LoRA

<Gallery />

## Model description 

This LoRA model specializes in generating portraits with crooked teeth, adding a unique and realistic touch to character images. Trained for 8 epochs over 1104 steps, it excels at producing yearbook-style portraits with varying degrees of dental misalignment.

The model aims to:
- Create diverse representations of crooked teeth
- Maintain natural facial expressions, particularly smiles

![A photo of a man with crooked teeth. He is wearing a brown jacket and a white shirt. The man has short brown hair and is wearing glasses. The background is blurred and contains a few objects.](https://cdn-uploads.huggingface.co/production/uploads/64a50a9a4f46b933c896a4ed/FUMs5aZDD1Xu0GxvausWZ.png)

This LoRA is particularly useful for:
- Character designers seeking to add distinctive features
- Artists looking to create more diverse and realistic portraits
- Projects requiring specific dental characteristics in images

While primarily focused on crooked teeth, the model maintains overall image quality and facial structure integrity.

## Trigger words

To activate this LoRA's specific features, include `crooked teeth` in your prompts.

## Download model

The weights for this model are available in Safetensors format.

You can [download](https://huggingface.co/Rejekts/Crooked-Teeth/resolve/main/crooked-teeth.safetensors) them in the Files & versions tab.

## Usage Tips

- Experiment with different strength settings to control the intensity of the crooked teeth effect
- Combine with other LoRAs or models for more diverse results
- Use negative prompts to fine-tune unwanted features if necessary

Remember, results may vary based on the base model and other parameters used in your generation pipeline.