AI Model: Aki (Character LoRA)

Base model: CyberRealistic Pony v14+
VAE: vae-ft-mse-840000-ema-pruned
Resolution: 768×1152
Trigger: aki_character

💫 Overview

Aki is an AI-based digital muse — bright, cinematic, and expressive.
She combines the softness of natural light with vivid emotional presence, adapting effortlessly between fashion, lifestyle, and cinematic portraiture.

Trained on a refined 9:16 dataset featuring portraits, mid-shots, and full-body compositions, she maintains stable anatomy, consistent lighting, and delicate facial balance.

Her aesthetic blends soft modern glow with subtle urban energy — denim, leather, pastel tops, and neon backdrops. Aki fits equally well in studio settings and stylized outdoor scenes.

🎨 Style & Use

Perfect for:

• cinematic and editorial portraits

• lifestyle or street-style photography

• creative and aesthetic storytelling

• stylized AI influencer or virtual model projects

• cosplay and stage aesthetics

• mature / 18+ creative projects (permitted)

Aki performs consistently across portrait, ¾, and full-body shots while preserving her unique facial identity and tone.

⚙️ Recommended Settings

Parameter Value

Weight 0.6–0.8

CFG 6–7

Steps 30–60

Sampler DPM++ 2M Karras

Resolution 768×1152 or 832×1216

VAE vae-ft-mse-840000-ema-pruned

Prompt example:

aki_character, cinematic portrait, soft daylight, realistic skin texture, gentle expression, sharp focus, natural lighting

Negative prompt:

blurry, low contrast, low detail, oversharpened, overexposed, distorted anatomy, flat light

🧠 Notes

• Works best with CyberRealistic Pony v14+.

• Balanced dataset ensures natural light behavior and realism.

• Handles soft focus, subtle bokeh, and mid-tone contrast beautifully.

• Suitable for editorial, lifestyle, and creative 18+ projects.

⚖️ Disclaimer

This LoRA was trained entirely on AI-generated imagery.
It does not depict a real person, nor is it based on any existing individual.
Aki is a fictional AI character created for artistic and creative purposes.
All resulting artworks generated with this LoRA belong fully to the users.