## Overview
Transform a single character image into a clean three-view character sheet.
This workflow generates:
* Front View
* Side View
* Rear View
from a single reference image while preserving the character's identity, style, clothing, and overall appearance.
Whether your reference is an AI-generated portrait, anime screenshot, game artwork, illustration, or original concept art, the workflow attempts to create a consistent multi-view reference suitable for further creative work.
---
# Features
✔ Character consistency across multiple views
✔ Generates Front / Side / Rear views
✔ Automatic character sheet composition
✔ Clean white background
✔ Minimal setup
✔ Low VRAM friendly
✔ Designed for Flux2.Klein GGUF
✔ Optimized for RTX 2080 (8 GB) and similar hardware
---
# Sample Workflow
```
Reference Image
│
▼
Character Extraction
│
▼
Front View Generation
│
├──────────────► Side View
│
└──────────────► Rear View
│
▼
Character Sheet Assembly
│
▼
Final Output
```
---
# Workflow Structure
The workflow is intentionally simple.
## 1. Load Character Reference
Load any character image.
Tested examples include:
* AI generated images
* Anime screenshots
* Donghua screenshots
* Concept art
* Illustrations
Portraits also work well.
The model will infer missing portions of the character where necessary.
---
## 2. Prompt Configuration
Prompting is separated into two logical layers.
### Global / Character Prompt
Contains instructions shared by every generation.
Examples:
* quality
* background
* consistency
* rendering style
* clothing
* accessories
* hairstyle
* identity
---
### View Prompt
Contains only view-specific instructions.
Examples:
Front
```
Generate a full-body front-facing view while maintaining character consistency.
```
Side
```
Generate a full-body 90 degree right side view while maintaining character consistency.
```
Rear
```
Generate a full-body back view while maintaining character consistency.
```
Keeping these prompts separate makes customization easier without affecting the entire workflow.
---
## 3. Character Sheet Generation
The workflow generates each view independently.
This is particularly useful for characters containing:
* tails
* wings
* capes
* backpacks
* armor
* oversized accessories
Each view can be tuned individually if required.
---
## 4. Character Sheet Assembly
The generated views are automatically combined into a single reference sheet.
Current layout:
```
Front
Side
Rear
```
---
# Supported Character Types
Successfully tested with:
✔ Realistic Humans
✔ Anime Characters
✔ Donghua Characters
✔ Anthropomorphic Characters
✔ Fantasy Characters
---
# Performance
Test Hardware
* NVIDIA RTX 2080 (8 GB)
* 64 GB System RAM
Typical generation time:
80–90 seconds (with 1024 longest side resolution cap)
for an entire character sheet.
Generation is performed sequentially to reduce VRAM usage.
---
# Notes
The workflow is designed to generate plausible multi-view references.
If the original image does not reveal the complete costume or body, the model intelligently infers the unseen portions.
As with any generative model, inferred details may not exactly match the original creator's intent.
Higher quality reference images generally produce better results.
---
# Tips
* Use high-quality reference images whenever possible.
* Portraits are supported, but full-body references may reduce ambiguity.
* If a character has unique back details (cape, wings, tail, tattoos, etc.), consider refining the Rear View prompt.
* For unusual poses or highly stylized characters, small prompt adjustments may improve consistency.
---
# Verified Model
This edition is designed and tested with:
Flux2.Klein (Q6 GGUF)
The workflow has been validated using the Q6 quantized model, which provides an excellent balance between quality, VRAM usage, and generation speed.
---
# Acknowledgements
The workflow orchestration is provided by the FisherKing Character Sheet Creator.
Character understanding, identity preservation, and inference of unseen details are made possible by the capabilities of the Flux2.Klein model.
---
# Final Thoughts
This workflow was created with a simple goal:
> Generate clean, consistent character reference sheets from a single image without requiring high-end hardware.
If you find it useful, I'd love to see the characters you create. Feel free to share your results, improvements, or remixes with the community. 🚀