ℹ️ Important information

What this LoRA is for: it turns fictional film and TV characters into collectible figures shown in a realistic setting: a sculpted, caricatured figure with a painted vinyl finish, standing full-body in a scene from its own story, shot from a low angle with a softly blurred background.

Fictional characters, not real people: the LoRA is meant to evoke characters: their iconic outfits, props, poses and settings. It is not intended for portraits of real people, and it must not be used to create images of real individuals.

What to expect: it was trained on one specific look and framing, a full-body figure with a scene behind it. It does that very well, but don't expect results that are very different from the examples in the gallery. Other compositions, close-up portraits or a photographic style are outside what it was made for.

How to prompt: realfig, <Character> from <Film or series>, full body, <outfit and scene> at weight 1, with Krea 2 Turbo at 8 steps and CFG 1.

Optimal resolution: 1280 px. The LoRA was trained at 1280, so generate at that size for the best detail and likeness.

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

Real Figurines

A style LoRA for Krea 2 that turns film and TV characters into collectible figures: sculpted, caricatured likeness, painted vinyl finish, low heroic camera angle and a blurred scene from their own story behind them.

Usage

Include the trigger word realfig in your prompt and name the character and the work they come from — the LoRA was trained that way, so it caricatures each character instead of giving everyone the same face:

realfig, <Character> from <Film or series>, full body, <outfit and scene>

Recommended weight: 1. Made for Krea 2 Turbo (8 steps, CFG 1); works with original characters too — just describe them.

Prompt Examples

"realfig, Tyrion Lannister in a red and gold doublet with a goblet of wine, full body, in a throne room"

"realfig, Sherlock from the BBC series in a long coat and blue scarf, full body, in a Baker Street flat"

"realfig, the Demogorgon from Stranger Things with a petal-shaped open head, full body, in a dark Upside Down forest"

"realfig, Rick Blaine from Casablanca in a white dinner jacket, full body, in a Moroccan nightclub"

"realfig, the Doctor from Doctor Who in a tweed jacket and bow tie with a sonic screwdriver, full body, in front of a blue police box"

"realfig, Hannibal Lecter in a white prison jumpsuit and face mask, on a hand truck, full body, in a stone prison cell corridor"

"realfig, Ace Ventura in a Hawaiian shirt with a giant quiff, full body, in a zoo"

"realfig, Thomas Shelby from Peaky Blinders in a flat cap and a long overcoat, full body, in a smoky industrial Birmingham street"

"realfig, Captain Kirk in his gold 1960s Starfleet tunic, full body, on an alien planet with red rocks"

"realfig, Chucky the killer doll in striped sweater and dungarees, full body, in a child's bedroom"

Training

• Base model: Krea 2 (trained on Raw)
• Type: Style LoRA
• Steps: 3000
• Network rank / alpha: 32 / 32
• Learning rate: 0.0002
• Optimizer: adamw8bit
• Resolutions: 1024 / 1280

Generation settings

Trained on Krea 2 Raw, and transfers to Krea 2 Turbo for faster generation.

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

The preview images were generated with krea2_turbo_mxfp8.safetensors, at 8 steps, guidance 1, LoRA weight 1.