Premium character LoRAs

Original synthetic characters, bundles and AI tools are in my shop:

https://rkrra.com/civitai

Get the character LoRAs on Patreon: https://www.patreon.com/cw/RkRRa

Join the community on Discord: https://discord.gg/nB79wRNtMT

Animal Kingdom

A style LoRA for Krea 2 that puts any animal right in front of the lens, as if it had walked up to the camera: an extreme ultra-wide fisheye a few centimeters from its face, a low worm's-eye angle from the ground, paws and claws reaching toward the viewer, and a huge landscape stretching far behind it. The light is always the same spectacle: a low sun in the frame with a starburst flare, golden HDR glow, dramatic clouds and saturated colors.

It works with mammals, birds, reptiles, fish, insects and sea creatures, and with several kinds of shot:

• From the ground, the animal leaning into the lens, like a selfie.
• Half underwater, the waterline splitting the frame between the reef and the sky.
• From below the surface, looking up through the water toward the sun.
• From inside a burrow, the animal peering down into the hole.
• In the air, wings spread as it swoops over the camera.
• Macro, with insects and frogs looming over giant dewdrops.

Usage

Start with the trigger word rkwildlens. The LoRA brings the light and the colors on its own, but describing the shot in the prompt is what makes it look like the gallery: the fisheye, how close the lens is, the angle, the foreground and the distant landscape.

rkwildlens, Extreme ultra-wide fisheye photo taken from <where>, the lens inches from the animal, low angle looking up. <Animal and what it is doing>, leaning into the lens. <Foreground>. Behind it, far below, <landscape>. A low sun sits in the frame with a bright starburst flare, golden HDR light, dramatic glowing clouds.

Recommended weight: 1.0. Made for vertical images (9:16); the training images were 941×1672.

Prompt Examples

"rkwildlens, Extreme ultra-wide fisheye photo taken from ground level, the lens only a few centimeters from the animal, low worm's-eye angle looking up. A cheetah sprinting straight at the camera, dust flying, leaning curiously into the lens with its face huge and close, one paw reaching out toward the camera. Dewy pebbles, moss and soil fill the blurred foreground. Behind it, far below, endless golden grassland and distant hills. A low sun sits in the frame with a bright starburst flare, golden HDR light, dramatic glowing clouds, rich saturated colors and crisp detail."

"rkwildlens, Extreme ultra-wide fisheye half-underwater split shot, the waterline crossing the middle of the frame, the lens right at the surface and inches from the animal. A sea otter floating on its back holding a shell. Below the waterline, clear turquoise water with bubbles and sun rays; above it, kelp beds and rocky sea stacks. A low sun sits in the frame with a bright starburst flare, golden HDR light, dramatic glowing clouds, rich saturated colors and crisp detail."

"rkwildlens, Extreme ultra-wide fisheye photo taken from inside a burrow, looking up through the round earthy opening. A spotted hyena, peering down into the hole straight at the lens, its face filling the opening. Roots and crumbs of soil frame the edge, and beyond it a blue sky with acacia trees. A low sun sits in the frame with a bright starburst flare, golden HDR light, dramatic glowing clouds, rich saturated colors and crisp detail."

"rkwildlens, Extreme ultra-wide fisheye photo taken from a branch, the lens inches from the animal, low angle looking up. A keel-billed toucan holding a red berry in its huge beak, leaning toward the camera with one limb reaching out to touch it. Leaves and dewdrops blur the foreground, and far below a jungle waterfall. A low sun sits in the frame with a bright starburst flare, golden HDR light, dramatic glowing clouds, rich saturated colors and crisp detail."

Training

• Base model: Krea 2
• Type: Style LoRA
• Steps: 5000
• Rank: 32
• Resolution: 1254 (native 941×1672)
• Optimizer: adamw, constant learning rate
• Engine: ai-toolkit

Generation settings

Trained on Krea 2, and works on Krea 2 Turbo for faster generation.

• Raw: ~30 steps, guidance 3.5
• Turbo: ~8 steps, guidance 1.0

The gallery images were generated with krea2_turbo_mxfp8.safetensors, at 8 steps, guidance 1, LoRA weight 1, 944×1672.