It is recommended to write the following tokens: many different flowers, blurry background, nebulous bouquets, liquid smoke, RAW, high quality, floating in the smoke, foggy, purple flowers in a blue vase, red flowers.
Copy the prompt and parameters from the preview to this model — you will not lose.
The model trained on photographs of flowers placed in liquid nitrogen vapors. This gave the effect of haze and fog, which the model remembered well.
LoRA was trained on a variety of examples, so try combinations and share your creativity in the comments.
I trained LoRA on the Deliberate_v3 model, so it's better to work with her in pairs.
I recommend using sd-dynamic-thresholding extension to increase the stability of generation and improve quality.
The model was trained for a resolution of 1024px, so you can generate huge images.
It is best to use the Deliberate_v3 model, but you can combine this LoRA with other solutions focused on photorealism, model will work perfectly with them.