Latest page update: Full Kontext support added, allowing contextual generation based on multiple visual and textual inputs. New recommended workflow featuring image → image → LoRA → prompt via Kontext. More consistent and creative results.

🧠 Recommended Local ComfyUI Workflow:

Download the JSON workflow here

Requirements:

• ComfyUI

• FLUX (or Kontext-compatible) base model

• Desired LoRA

• Kontext plugin enabled

⚙️ How It Works:

This workflow is designed to operate on 4 input layers:

1. A base image, used as the structural and visual foundation.

2. A reference image, used for style, pose, or extra inspiration.

3. A custom LoRA, to inject aesthetic, character, or specific style.

4. A text command, interpreted by Kontext, that intelligently guides and contextualizes the final result.

The model treats the user's "command" as part of the image's narrative universe, making the output more coherent, responsive, and often surprising.

🪄 Trigger (must be included in the prompt for best results):

contextual command: + whatever you want (e.g. a character in a cyberpunk favela watching the sunrise with his crew)

Example usage in the prompt:

less

CopyEdit

contextual command: a chavoso time traveler with grills and dreads looking at medieval ruins through AR glasses

⚠️ Important:

Kontext must be enabled and linked to FLUX (or another compatible model).
Running this locally in ComfyUI is strongly recommended, as it allows the use of custom samplers, schedulers, and advanced controls that are not available in online generators.

🔍 What Makes This Workflow Special?

• Merges aesthetic + reference + verbal command into a continuous, unified pipeline

• Avoids generic or out-of-context generations

• Works great with original characters, visual mashups, and narrative-driven scenes

• Understands the relationship between image and text, unlike most basic workflows

🧬 LoRA Notes (if you're using a specific LoRA):

This LoRA was trained to maintain stylistic consistency without sacrificing the originality of the prompt. When combined with Kontext, it dynamically adapts to the user's command, balancing visuals and narrative intent.

⚖️ Disclaimer:

This model does not aim to accurately or legally represent any real person, copyrighted concept, or protected likeness. All generated content is fictional and experimental. If you are the rightful owner of any likeness or idea you believe is being emulated and want it removed, please contact me directly with proof of ownership.

This model is not intended for illegal use or morally questionable activities like spreading misinformation through deepfakes.

Let me know if you want a version tailored for Civitai’s actual submission form (with model type, tags, etc.), or a layout with visuals like the ComfyUI node tree. Just say the word: Let’s get it.