Triggers: Pelican

Model Strength: 0.8–1 (dependent on the generator)

Usage Tips:

Captioning data often included, phrases like "A high-definition image of a pelican" or "A high-quality image of a pelican" these can enhance the triggering of the LoRA. Specifying the exact species of pelican usually produces accurate variations in feathers and markings, although some adjustments may be needed to achieve perfection. Examples include:

• American White Pelican

• Brown Pelican

• Dalmatian Pelican

• Great White Pelican

• Pink-backed Pelican

• Australian Pelican

• Spot-billed Pelican

• Peruvian Pelican

I have been using Flux for a while now, and during this time, I’ve encountered a significant issue with it: pelicans often appear as a cross between a heron and a pelican. The sizes are inconsistent, beaks are too thin and sometimes feature a strange ball at the tip instead of the characteristic curve, and the feet resemble claws rather than webbed totipalmate feet. Overall, the pelicans lack the rotund gracefulness they are known for and appear runty and unrefined.

This LoRA aims to address these shortcomings by improving the quality of the feathers, refining the beak shapes, ensuring webbed feet, and allowing users to specify the pelican species in the prompt.

Areas for Improvement:
One area where this LoRA still struggles is in generating accurate flight images. While pelicans often emerge with their heads tucked properly, some still display the S-shaped neck posture typical of herons. Despite these challenges, most results look significantly better with this improvement.