lbc_f.1_dev_Antiquities

This model follows the Flux-1-dev open-source license and is intended solely for educational and research purposes.

It has been trained on the original and complete f.1-dev version using a dataset of over 600 carefully curated Chinese-style photography collections, along with an additional training set of East Asian facial features. And, of course, it also includes the elements you’ve been eagerly anticipating ****.

The model was trained on an NVIDIA A800 80G for approximately 23 hours, and testing indicates that it meets the preferences of most Chinese-style photography enthusiasts.

The diverse sample images demonstrate that the model can effectively achieve desired effects when combined with various stylized Lora models. This advantage stems from not integrating other foundational models or Lora models during the fine-tuning process, thus preserving the original base model's pure encoding information. In other words, this model can serve as a foundational base for Lora training.