controlnet-lllite-xl-pd-canny-alpha

ControlNet LLLite XL PD Canny Alpha

Overview

ControlNet-LLLite Weights trained on stabilityai/stable-diffusion-xl-base-1.0 and LEOSAM's HelloWorld XL with canny conditioning.

This model was trained to generate product photography images with a pure white background. You can use it with prompt words like 'pure white background', 'plain white background' to generate product photography style images with a fairly stable white background.

This model is just an experiment and is not yet stable.

The generated results seems can be significantly influenced by the base model. This might be due to the fact that 'white background' is a common prompt word. As a result, it is not recommended to choose models other than stabilityai/stable-diffusion-xl-base-1.0 and LEOSAM's HelloWorld XL as the base model, to ensure better results.

Here are some comparisons with other models:

Compared to other models including stabilityai/control-lora, diffusers/controlnet-canny-sdxl-1.0, and kohya-ss/controlnet-lllite, our model tends to generate images with pure white backgrounds under studio lighting conditions. Among them, diffusers/controlnet-canny-sdxl-1.0 has the best control ability. stabilityai/control-lora and kohya-ss/controlnet-lllite each have their unique characteristics in terms of style and control. Users can choose according to their specific needs.

Naming Rule

controlllite_xl_pd_canny_alpha_sdxl_020.safetensors

• sdxl: base model
• 020: num of epoch