gymBall-Lora-v1.0

gymBall Lora V1.0

<Gallery />

Model description

Gym Ball LoRA – User Guide

This LoRA specializes in realistic interaction between a character and an exercise/gym ball. It focuses on:

• Accurate ball texture and surface (matte, satin, glossy, pebbled vinyl)
• Correct ball size relative to the body
• Realistic deformation when weight is applied (from fully round to heavily squashed)
• Natural body-ball contact and posing (sitting, straddling, leaning, resting on the ball, etc.)

The main strength is how the ball compresses and bulges under the character’s weight, along with proper contact points and pose.

Trigger word: `gymBall`

Key Controllable Terms

Framing (pick exactly one):
`extreme close-up` / `close-up` / `medium shot` / `full body shot` / `wide shot`

Ball size (pick one):
`small exercise ball` / `standard exercise ball` / `large exercise ball` / `oversized exercise ball`

Surface finish (use these exact phrases):

• `matte anti-burst vinyl`
• `satin vinyl with a soft broad highlight`
• `glossy vinyl with a sharp specular highlight`
• `lightly pebbled vinyl`

Deformation (this is the most important control — choose one):

• `fully round and firmly inflated`
• `slightly compressed`
• `moderately compressed`
• `heavily compressed`
• `squashed into a wide flattened ellipsoid`

Tips for Best Results

• Start with LoRA weight 0.7 – 1.0. Higher weights increase deformation accuracy but can overpower the base model.
• Be explicit about both the deformation level and where the body contacts the ball.
• The LoRA responds strongly to the exact deformation phrases listed above — use them for consistent control.
• Negative prompt suggestions: `flat ball, undeformed, perfectly round when weight is applied, floating, deformed anatomy, extra limbs`
• Works well stacked with character or clothing LoRAs at moderate weights.

This LoRA is specialized for realistic ball shape, surface, and body-induced deformation. Keeping prompts focused on those elements produces the cleanest results.

License

This LoRA is a derivative of Krea 2 Raw and is released under the Krea 2 Community License.

You may use, modify, and distribute this LoRA.
Commercial use is allowed only if you (and any affiliated entities) have total annual revenue under $1,000,000 USD.
If your revenue is $1M or higher, you need an enterprise license from Krea.
You must also comply with Krea’s Acceptable Use Policy (no CSAM, no non-consensual intimate imagery of real people, etc.).

Full license text: https://github.com/krea-ai/krea-2/blob/main/docs/KREA-2-COMMUNITY-LICENSE

Trigger words

You should use gymBall to trigger the image generation.

Download model

Download them in the Files & versions tab.