flux cinestill

Introduction

FLUX CINESTILL is a LoRA fine-tuned model based on the FLUX.1-Dev base model. It is designed for generating text-to-image outputs that resemble night and dusk time photographs with a distinctive slight halation effect, reminiscent of Cinestill 800T film.

Architecture

The model is fine-tuned using the FLUX.1-Dev architecture, leveraging the capabilities of the Diffusers library for text-to-image generation. It employs LoRA (Low-Rank Adaptation) techniques to enhance the generation of specific styles, particularly those mimicking analog film characteristics.

Training

FLUX CINESTILL was trained using the Replicate platform, specifically focusing on Cinestill 800T images. The training process aimed to optimize the generation of high-resolution, film-like images, with keywords such as "cinestill 800t," "night time," "dusk," "4k," and "analog film" enhancing the model's output quality.

Guide: Running Locally

To run the model locally, follow these steps:

1. Install Dependencies: Ensure you have Python and PyTorch installed. Install the Diffusers library using pip:

   pip install diffusers
   

2. Load the Model: Use the following code to load the model and generate images:

   from diffusers import AutoPipelineForText2Image
   import torch
   
   pipeline = AutoPipelineForText2Image.from_pretrained(
       'black-forest-labs/FLUX.1-dev', 
       torch_dtype=torch.float16
   ).to('cuda')
   pipeline.load_lora_weights('adirik/flux-cinestill', weight_name='lora.safetensors')
   image = pipeline('your prompt').images[0]
   

   Replace 'your prompt' with a suitable prompt using the trigger word CNSTLL.

3. Cloud GPUs: For better performance, especially with high-resolution images, consider using cloud GPU services such as AWS, Google Cloud, or Azure.

License

FLUX CINESTILL is released under the flux-1-dev-non-commercial-license. For details, refer to the license link. This license permits non-commercial use, with all rights reserved.