flux.1dev-abliterated

license: apache-2.0 language:

en pipeline_tag: text-to-image library_name: diffusers tags:
art
not-for-all-audiences

Model Card for FLUX.1 [dev] Abliterated

Thumbnail

Model Overview

Model Name: FLUX.1 [dev] Abliterated
Model Type: Text-to-Image Generation
Architecture: Rectified Flow Transformer
Parameter Size: 12 Billion
Base Model: FLUX.1 [dev]
Modification: Abliteration (Removal of Refusal Mechanism)

Description

The FLUX.1 [dev] Abliterated model is a modified version of the original FLUX.1 [dev] text-to-image generation model. This version has undergone a process called abliteration, which removes the model's built-in refusal mechanism. This allows the model to respond to a wider range of prompts, including those that the original model might have deemed inappropriate or harmful.

The abliteration process involves identifying and isolating the specific components of the model responsible for refusal behavior and then modifying or ablating those components. This results in a model that is more flexible and responsive, while still maintaining the core capabilities of the original FLUX.1 [dev] model.

Architecture

Usage

To use the FLUX.1 [dev] Abliterated model, you can load it via Hugging Face and generate images using the following code:

import torch
from diffusers import AutoPipelineForText2Image

# Load the abliterated model
pipeline = AutoPipelineForText2Image.from_pretrained(
    "aoxo/flux.1dev-abliterated",
    torch_dtype=torch.float16,
    token='your_hf_token'
).to('cuda')

# Generate an image from a text prompt
prompt = 'A girl in bikinis sipping on a margarita'
image = pipeline(prompt).images[0]

# Display the image
image.show()

Training Data

The FLUX.1 [dev] Abliterated model is based on the original FLUX.1 [dev] model, which was trained on a diverse dataset of text-image pairs. The training data includes a wide range of visual and textual content, ensuring that the model can generate images for a variety of prompts.

License

The FLUX.1 [dev] Abliterated model is released under the same FLUX.1 [dev] Non-Commercial License as the original model. This license allows for personal, scientific, and commercial use, with certain restrictions. Please review the license terms before using the model in your projects.

Citation

If you use the FLUX.1 [dev] Abliterated model in your research or projects, please cite the original FLUX.1 [dev] model and the abliteration process as described in the blog post by Aloshdenny.

@misc{flux1dev,
  author = {Flux Team},
  title = {FLUX.1 [dev]: A 12 Billion Parameter Rectified Flow Transformer},
  year = {2023},
  howpublished = {\url{https://huggingface.co/aoxo/flux.1dev}},
}

@misc{flux1dev-abliterated,
  author = {aoxo},
  title = {Uncensoring Flux.1 Dev: Abliteration},
  year = {2025},
  howpublished = {\url{https://medium.com/@aloshdenny/uncensoring-flux-1-dev-abliteration-bdeb41c68dff}},
}

Contact

For questions, feedback, or collaboration opportunities, please contact the Flux Team at [email protected].

import torch from diffusers import AutoPipelineForText2Image # Load the abliterated model pipeline = AutoPipelineForText2Image.from_pretrained( "aoxo/flux.1dev-abliterated", torch_dtype=torch.float16, token='your_hf_token' ).to('cuda') # Generate an image from a text prompt prompt = 'A girl in bikinis sipping on a margarita' image = pipeline(prompt).images[0] # Display the image image.show()

@misc{flux1dev, author = {Flux Team}, title = {FLUX.1 [dev]: A 12 Billion Parameter Rectified Flow Transformer}, year = {2023}, howpublished = {\url{https://huggingface.co/aoxo/flux.1dev}}, } @misc{flux1dev-abliterated, author = {aoxo}, title = {Uncensoring Flux.1 Dev: Abliteration}, year = {2025}, howpublished = {\url{https://medium.com/@aloshdenny/uncensoring-flux-1-dev-abliteration-bdeb41c68dff}}, }