https://www.reddit.com/r/StableDiffusion/comments/1lx8lot/kontext_presets_all_system_prompts/

https://huggingface.co/QuantStack/FLUX.1-Kontext-dev-GGUF

https://huggingface.co/mit-han-lab/nunchaku-flux.1-kontext-dev

https://docs.comfy.org/tutorials/flux/flux-1-kontext-dev

If you require high fidelity in image details, please avoid using the following nodes:

https://github.com/Zehong-Ma/ComfyUI-MagCache

https://github.com/ChenDarYen/ComfyUI-NAG

## 🔤 Prompt Verb Reference Table

| **Verb**    | **Purpose**                                     | **Example Prompt**                                                                 |

|-------------|-------------------------------------------------|------------------------------------------------------------------------------------|

| `Add`       | Introduce new elements                          | Add ASCII style text only the single word "In", no additional letters              |

| `Change`    | Modify attributes or background                 | Change the background to a beach, keep the subject in the exact same position and pose |

| `Convert`   | Transform visual form, style, or material       | Convert to pencil sketch with natural graphite lines, cross-hatching, and visible paper texture |

| `Move`      | Relocate subject or object                      | Move the dog to the left side, keeping its size and pose unchanged                 |

| `Preserve`  | Explicitly define elements to remain untouched  | Preserve facial expression and body position                                       |

| `Reduce`    | Decrease intensity, size, or complexity         | Reduce brightness while keeping color palette and composition unchanged            |

| `Remove`    | Delete objects or features                      | Remove all background people while preserving the subject's position and lighting  |

| `Replace`   | Swap one element for another                    | Replace 'joy' with 'BFL', maintain the same font style                             |

| `Rotate`    | Turn camera or object perspective               | Rotate the camera 180 degrees to view from behind                                  |

| `Switch`    | Change orientation or view                      | Switch to side profile while keeping facial features unchanged                     |

| `Turn`      | Reorient subject's body/facing direction        | Turn the subject to face forward completely, including both head and body          |

| `Transform` | Overhaul style or scene context                 | Transform to 1960s pop art style with bright colors and commercial graphics        |

---

## ✅ Best Practice Templates with Pitfalls

### 1. Basic Modifications

- ❌ *Incorrect:* "Make the car red"  

- ✅ *Better:* "Change the car color to red while keeping the environment unchanged"

### 2. Style Transfer

- ❌ *Incorrect:* "Make it a sketch"  

- ✅ *Better:* "Convert to pencil sketch with natural graphite lines, cross-hatching, and visible paper texture"

### 3. Character Consistency

- ❌ *Incorrect:* "Transform the person into a Viking"  

- ✅ *Better:* "Change the clothes to be a Viking warrior while preserving facial features"

### 4. Composition Control

- ❌ *Incorrect:* "Put her on a beach"  

- ✅ *Better:* "Change the background to a beach while keeping the person in the exact same position and pose"

### 5. Text Editing

- ❌ *Incorrect:* "Write BFL instead of joy"  

- ✅ *Better:* "Replace 'joy' with 'BFL', maintain the same font style"

### 6. View Switching

- ❌ *Incorrect:* "Make it a front view"  

- ✅ *Better:* "Turn the subject to face forward completely, including both head and body, while maintaining the same outfit, pose, and composition"

### 7. Perspective Rotation

- ❌ *Incorrect:* "Show from behind"  

- ✅ *Better:* "Rotate the camera 180 degrees to view from behind while keeping the 3D style and lighting consistent"

### 8. Background Adjustment

- ❌ *Incorrect:* "Add indoor background"  

- ✅ *Better:* "Replace the background with a bright indoor room while preserving lighting, subject position, and scale"

### 9. Lighting & Atmosphere

- ❌ *Incorrect:* "Make it sunny"  

- ✅ *Better:* "Change to an early morning scene with soft golden sunlight and light mist, keeping architecture unchanged"

---

## 🧠 Core Prompting Principles

1. **Be Specific and Clear** – Use exact terms to define action and preserved elements

2. **Step-by-step Editing** – Separate large edits into logical, incremental changes

3. **Explicit Preservation** – Always declare what must remain unchanged (pose, lighting, expression, etc.)

4. **Action Verbs First** – Start prompts with strong, direct verbs (`change`, `replace`, `convert`, etc.)

5. **No Ambiguity** – Avoid vague instructions like “make it better” or “transform it” without context

> FLUX.1 Kontext performs best when the prompt is precise, concrete, and directive.