Holly Face v2.0 — Consistent AI Character LoRA for FLUX.2 Klein 9B

Overview

Holly Face v2.0 is a LoRA (Low-Rank Adaptation) designed to produce a consistent, recognizable original character across a wide range of prompts, poses, lighting conditions, and artistic styles. This LoRA was trained specifically for the FLUX.2 Klein 9B model and produces the best results when used with that base model.

The character "Holly" is an original digital creation — a young woman with distinctive, recognizable facial features that remain consistent across every generation.

Trigger Word

h0lly

Important: Use the number zero (0), not the letter "o". This trigger word must be included in your prompt for the LoRA to activate.

Recommended Generation Settings

| Setting | Value |

|---|---|

| Base Model | FLUX.2 Klein 9B |

| CFG Scale | 4.0 |

| Steps | 4 |

| Sampler | Euler |

| Scheduler | Normal |

| LoRA Weight | 0.85 |

| Resolution | 768×1024, 1024×1024 |

| Negative Prompt | None needed |

How to Use

Include the trigger word h0lly in your prompt along with any scene, style, or mood description. The LoRA handles face consistency — you control everything else (lighting, pose, background, clothing, etc.) through your prompt text.

Example Prompts

Portrait:

h0lly, soft studio lighting portrait, gentle smile, looking directly at camera, photorealistic, high detail skin texture

Outdoor:

h0lly close-up portrait, natural golden hour sunlight, outdoor garden with flowers, cinematic depth of field, warm tones

Professional:

h0lly, professional headshot, soft rim lighting from behind, slight turn to the left, confident expression, dark gradient background, corporate photography style

Full Body:

h0lly full body, standing in a modern art gallery, wearing a black dress, arms relaxed at sides, wide angle shot, interior lighting

Artistic:

h0lly, oil painting style portrait, dramatic chiaroscuro lighting, renaissance composition, rich warm colors, visible brush strokes

Casual:

h0lly, casual selfie angle, slight tilt of head, natural daylight, wearing a hoodie, relaxed expression, slightly desaturated tones

LoRA Weight Guide

| Weight | Effect |

|---|---|

| 0.5 – 0.7 | Lighter character influence, more creative flexibility, face may drift |

| 0.7 – 0.85 | Balanced — good consistency while allowing prompt control (recommended) |

| 0.85 – 1.0 | Strong character lock, maximum face consistency, less prompt flexibility |

Training Details

- Training Platform: Civitai On-Site Trainer

- Base Model: FLUX.2 Klein 9B Base

- Epochs: 10

- Learning Rate: 0.0001

- LR Scheduler: Cosine

- Network Dim: 32

| Network Alpha: 32

| Optimizer: adamw8bit

- Training Images: Multiple poses including front-facing, side profile, three-quarter angle, close-up, and full body

- Captions: Detailed pose-specific captions for each training image

Tips for Best Results

1. Always include h0lly at the start of your prompt for strongest activation

2. Use FLUX.2 Klein 9B as your base model — results on other FLUX variants may vary

3. Keep steps at 4 for speed, or increase to 8–12 for slightly more detail

4. A LoRA weight of 0.85 is the sweet spot for most use cases

5. Be descriptive with lighting and style — the LoRA handles the face, you handle the scene

6. No negative prompt is required with FLUX.2 Klein 9B

License

This LoRA is released for personal and creative use. The character Holly is an original creation. Please do not use this LoRA to generate misleading or deceptive content.