ComfyUI Flux Kontext Dev Native Workflow Example

ComfyUI Flux Kontext Dev Native Workflow Example.

About FLUX.1 Kontext Dev

FLUX.1 Kontext is a breakthrough multimodal image editing model from Black Forest Labs that supports simultaneous text and image input, intelligently understanding image context and performing precise editing. Its development version is an open-source diffusion transformer model with 12 billion parameters, featuring excellent context understanding and character consistency maintenance, ensuring that key elements such as character features and composition layout remain stable even after multiple iterative edits.

It shares the same core capabilities as the FLUX.1 Kontext suite:

• Character Consistency: Preserves unique elements in images across multiple scenes and environments, such as reference characters or objects in the image.
• Editing: Makes targeted modifications to specific elements in the image without affecting other parts.
• Style Reference: Generates novel scenes while preserving the unique style of the reference image according to text prompts.
• Interactive Speed: Minimal latency in image generation and editing.

While the previously released API version offers the highest fidelity and speed, FLUX.1 Kontext [Dev] runs entirely on local machines, providing unparalleled flexibility for developers, researchers, and advanced users who wish to experiment.

Version Information

• [FLUX.1 Kontext [pro] - Commercial version, focused on rapid iterative editing
• FLUX.1 Kontext [max] - Experimental version with stronger prompt adherence
• FLUX.1 Kontext [dev] - Open source version (used in this tutorial), 12B parameters, mainly for research

Currently in ComfyUI, you can use all these versions, where Pro and Max versions can be called through API nodes, while the Dev open source version please refer to the instructions in this guide.

Workflow Description

In this tutorial, we cover two types of workflows, which are essentially the same:

• A workflow using the FLUX.1 Kontext Image Edit group node, making the interface and workflow reuse simpler
• Another workflow without using group nodes, showing the complete original workflow.

The main advantage of using group nodes is workflow conciseness - you can reuse group nodes to implement complex workflows and quickly reuse node groups. Additionally, in the new version of the frontend, we’ve added a quick group node addition feature for Flux.1 Kontext Dev:

This feature is currently experimental and may be adjusted in future versions.

Make sure your ComfyUI is updated.

• Download ComfyUI
• Update Guide

Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.(Desktop version’s update will delay sometime)

If nodes are missing when loading a workflow, possible reasons:

1. Not using the latest ComfyUI version(Nightly version)
2. Using Stable or Desktop version (Latest changes may not be included)
3. Some nodes failed to import at startup

Model Download

To run the workflows in this guide successfully, you first need to download the following model files. You can also directly get the model download links from the corresponding workflows, which already contain the model file download information.

Diffusion Model

• flux1-dev-kontext_fp8_scaled.safetensors

If you want to use the original weights, you can visit Black Forest Labs’ related repository to obtain and use the original model weights.

VAE

• ae.safetensors

Text Encoder

• clip_l.safetensors
• t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn_scaled.safetensors

Model save location

📂 ComfyUI/
├── 📂 models/
│   ├── 📂 diffusion_models/
│   │   └── flux1-dev-kontext_fp8_scaled.safetensors
│   ├── 📂 vae/
│   │   └── ae.safetensor
│   └── 📂 text_encoders/
│       ├── clip_l.safetensors
│       └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn_scaled.safetensors

Flux.1 Kontext Dev Workflow

This workflow is a normal workflow, but it uses the Load Image(from output) node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing.

1. Workflow and Input Image Download

Download the following files and drag them into ComfyUI to load the corresponding workflow

Input Image

2. Complete the workflow step by step

You can refer to the numbers in the image to complete the workflow run:

1. In the Load Diffusion Model node, load the flux1-dev-kontext_fp8_scaled.safetensors model
2. In the DualCLIP Load node, ensure that clip_l.safetensors and t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn_scaled.safetensors are loaded
3. In the Load VAE node, ensure that ae.safetensors model is loaded
4. In the Load Image(from output) node, load the provided input image
5. In the CLIP Text Encode node, modify the prompts, only English is supported
6. Click the Queue button, or use the shortcut Ctrl(cmd) + Enter to run the workflow

Flux Kontext Prompt Techniques

1. Basic Modifications

• Simple and direct: "Change the car color to red"
• Maintain style: "Change to daytime while maintaining the same style of the painting"

2. Style Transfer

Principles:

• Clearly name style: "Transform to Bauhaus art style"
• Describe characteristics: "Transform to oil painting with visible brushstrokes, thick paint texture"
• Preserve composition: "Change to Bauhaus style while maintaining the original composition"

3. Character Consistency

Framework:

• Specific description: "The woman with short black hair" instead of “she”
• Preserve features: "while maintaining the same facial features, hairstyle, and expression"
• Step-by-step modifications: Change background first, then actions

4. Text Editing

• Use quotes: "Replace 'joy' with 'BFL'"
• Maintain format: "Replace text while maintaining the same font style"

Common Problem Solutions

Character Changes Too Much

❌ Wrong: "Transform the person into a Viking" ✅ Correct: "Change the clothes to be a viking warrior while preserving facial features"

Composition Position Changes

❌ Wrong: "Put him on a beach" ✅ Correct: "Change the background to a beach while keeping the person in the exact same position, scale, and pose"

Style Application Inaccuracy

❌ Wrong: "Make it a sketch" ✅ Correct: "Convert to pencil sketch with natural graphite lines, cross-hatching, and visible paper texture"

Core Principles

1. Be Specific and Clear - Use precise descriptions, avoid vague terms
2. Step-by-step Editing - Break complex modifications into multiple simple steps
3. Explicit Preservation - State what should remain unchanged
4. Verb Selection - Use “change”, “replace” rather than “transform”

Best Practice Templates

Object Modification: "Change [object] to [new state], keep [content to preserve] unchanged"

Style Transfer: "Transform to [specific style], while maintaining [composition/character/other] unchanged"

Background Replacement: "Change the background to [new background], keep the subject in the exact same position and pose"

Text Editing: "Replace '[original text]' with '[new text]', maintain the same font style"

Remember: The more specific, the better. Kontext excels at understanding detailed instructions and maintaining consistency.