The model was created through a large-scale post-training of the pre-trained weights provided by Black Forest Labs. The post-training work was made possible by the hard work of the data, infrastructure, and product team who put together a solid foundation for the post-training pipelines. The model is capable of generating high-quality images with superior aesthetic control. The recommended inference settings include a resolution between 1024 and 1280 pixels, 28-32 inference steps, and CFG guidance between 3.5 and 5.0.
FLUX.1 Krea [dev] is a powerful tool for generating images with high aesthetic quality. The model is fully compatible with the FLUX.1 [dev] architecture and can be used for a variety of applications. The repository includes a live demo and a Jupyter Notebook to run the model, making it easy to get started. The model is also capable of generating images with custom prompts, allowing for a high degree of control over the output. The model is a valuable resource for anyone looking to generate high-quality images with superior aesthetic control.