Realistic Flux Workflow

[FULL GUIDE]

Mar 19, 2025

by Mickmumpitz

Over the last few months I created this powerful Flux workflow that produces exceptionally realistic images with remarkable consistency. This workflow works seamlessly with character Loras created using my Consistent Character Creator.

My workflow uses a combination of selected Loras, the use of negative prompts, and the detail Daemon sampler to achieve detailed photorealistic results that stand out from standard generations. The workflow also fully integrates with ControlNets, giving you complete authority over character positioning and overall composition.

Whether you're creating portraits, scenes, or complex character interactions, this workflow provides the precision and quality needed for professional-grade image generation.

1 Installation:

Download the .json file and drag and drop it into your ComfyUI window.

Install the missing custom nodes via the manager and restart ComfyUI.

Before You Begin:


Thank you for considering supporting us! Since these workflows can be complex, we recommend testing the free versions first to ensure compatibility with your system. We cannot guarantee full compatibility with every system that's why we always provide the main functionalities for free!


Please take a moment to read through the entire guide. If you encounter any issues:


  1. Check the troubleshooting section at the end of the guide

  2. If problems persist, visit our Discord's #need-help channel and use the search function—many common issues have already been resolved

  3. If you cannot find it there ask and we will try to help you. give as much information as possible. Include screenshots, the error.

2 Download the Models:

1. FLUX CHECKPOINT: flux1-dev-fp8
📁 ComfyUI_windows_portable\ComfyUI\models\checkpoints
Search for "Flux" in the model manager and install "flux1-dev-fp8".

2. CONTROL NET: InstantX/FLUX.1-dev Controlnet (Union)
📁 Open the ComfyUI Manager > Model Manager > search for “union” > install

3. UPSCALE MODEL: 4x-ClearRealityV1 https://openmodeldb.info/models/4x-ClearRealityV1*

📁 \ComfyUI_windows_portable\ComfyUI\models\upscale_models

* You can also use other upscale models like 4xUltraSharp which you can download via the ComfyUI Model Manager.

4. LORAS

I use a combination of 4 Loras:

My Character Lora that I trained with my Consistent Character Creator.

  1. Amateur Photography
    Strength: 0.6
    Download here: https://civitai.com/models/652699/amateur-photography-flux-dev

  2. UltraRealistic Lora Project
    Strength: 0.6
    Download: https://www.reddit.com/r/StableDiffusion/comments/1eywnv8/realism_comparison_v2_amateur_photography_lora/

  3. Boring Reality
    Strength: 0.8
    Download: https://civitai.com/models/639937/boreal-fd-boring-reality-flux-dev-lora

Make sure to play around with the strengths!

3 Workflow-Walkthrough

Prompt

After you downloaded and loaded all the models it is time to create the prompt. Be as detailed and descriptive as possible - Flux really likes that. Name the style, lighting, composition, action, look and feel of your image.

photography of a woman called a1mber standing in an italian village. she is wearing a grey turtleneck sweater, brunette hair, she is standing on an old bridge with views of a landscape in tuscany, candid photography, she is smiling. amateur photography with motion blur, iphone photo, hard sunlight, low resolution candid photo taken on a phone, grainy, the old rustic buildings are decorated in great detail, hard evening sunlight reflects from her skin

a1mber is the trigger word for my character lora!

ControlNet

A ControlNet is a tool that lets you guide AI image generation by using reference images or sketches to control specific aspects like pose, composition, or layout while still maintaining creative freedom in other elements.

Using it with this workflow is completely optional. Activate it using the fast bypasser in the beginning of the workflow.

I recommend using openpose or canny as inputs. I mostly use this online tool to pose my characters and then download the poses from there. Works really great most of the time, but don’t expect it to work every time. Sometimes you need to try out different seeds or increase the strength parameter in the “Apply Advanced ControlNet” node.

Sampler

  1. Set your scene dimensions here.

  2. The Detail Daemon Sampler is an insanely powerful tool to add detail to your generated images. I recommend checking out some tutorials for this tool alone, but make sure to play around with the detail_amount. I personally keep it between 0.10 - 0.60 since higher values can break the images but make sure to try it out yourself!

  3. batch_size: How many images you want to generate in one go.

  4. Noise seed for the image. Change it to get a completely new variation of your prompt. Set it to “randomize”

  5. These are the basic sampler settings and these work well for most images I generated. But feel free to play around with them and test out different sampler and scheduler combinations.

  6. Should be greater than 1. This value worked great for me but feel free to try out different values.

  7. This will save out a low res preview of your image. Deactivate this node using CTRL+B if you don’t need it.

Upscale

Here I am using an SDUltimate upscaler with Flux. This one will break down the image into smaller tiles and upscale them one after another creating more detail but also make it easier to process for your computer.

←I am using this combination of settings and I recommend you use them as a basis, but as always play around with them!

The most important value is the denoise value. This will change how much the upscaler is allowed to change your image. A value of 0.30 is a good starting point.

It will allow the sampler to change the image 30% giving it some freedom to change the image but also making sure that it stays consistent.

Post Processing

  1. Chromatic aberration is the visual distortion that occurs when a lens fails to focus all colors to the same point, causing colored fringes along boundaries between dark and light areas of an image.

  2. Image sharpening in phone photography is the process of enhancing edge contrast to make details appear more defined and crisp to compensate for limitations in small camera sensors.

  3. Bloom in phone photography is the halo-like glow that appears around bright light sources. Smudges on a phone lens can create or exaggerate bloom effects because fingerprints, oils, or dirt particles scatter incoming light in unpredictable directions.

  4. Film grain in phone photography is the deliberate addition of noise or texture to digital images to simulate the organic, slightly imperfect look of traditional film photographs.

  5. Image compression in phone photography is the process of reducing file sizes by selectively discarding visual data, often resulting in artifacts like blocky textures or loss of fine detail when overdone.

Over the last few months I created this powerful Flux workflow that produces exceptionally realistic images with remarkable consistency. This workflow works seamlessly with character Loras created using my Consistent Character Creator.

My workflow uses a combination of selected Loras, the use of negative prompts, and the detail Daemon sampler to achieve detailed photorealistic results that stand out from standard generations. The workflow also fully integrates with ControlNets, giving you complete authority over character positioning and overall composition.

Whether you're creating portraits, scenes, or complex character interactions, this workflow provides the precision and quality needed for professional-grade image generation.

1 Installation:

Download the .json file and drag and drop it into your ComfyUI window.

Install the missing custom nodes via the manager and restart ComfyUI.

Before You Begin:


Thank you for considering supporting us! Since these workflows can be complex, we recommend testing the free versions first to ensure compatibility with your system. We cannot guarantee full compatibility with every system that's why we always provide the main functionalities for free!


Please take a moment to read through the entire guide. If you encounter any issues:


  1. Check the troubleshooting section at the end of the guide

  2. If problems persist, visit our Discord's #need-help channel and use the search function—many common issues have already been resolved

  3. If you cannot find it there ask and we will try to help you. give as much information as possible. Include screenshots, the error.

2 Download the Models:

1. FLUX CHECKPOINT: flux1-dev-fp8
📁 ComfyUI_windows_portable\ComfyUI\models\checkpoints
Search for "Flux" in the model manager and install "flux1-dev-fp8".

2. CONTROL NET: InstantX/FLUX.1-dev Controlnet (Union)
📁 Open the ComfyUI Manager > Model Manager > search for “union” > install

3. UPSCALE MODEL: 4x-ClearRealityV1 https://openmodeldb.info/models/4x-ClearRealityV1*

📁 \ComfyUI_windows_portable\ComfyUI\models\upscale_models

* You can also use other upscale models like 4xUltraSharp which you can download via the ComfyUI Model Manager.

4. LORAS

I use a combination of 4 Loras:

My Character Lora that I trained with my Consistent Character Creator.

  1. Amateur Photography
    Strength: 0.6
    Download here: https://civitai.com/models/652699/amateur-photography-flux-dev

  2. UltraRealistic Lora Project
    Strength: 0.6
    Download: https://www.reddit.com/r/StableDiffusion/comments/1eywnv8/realism_comparison_v2_amateur_photography_lora/

  3. Boring Reality
    Strength: 0.8
    Download: https://civitai.com/models/639937/boreal-fd-boring-reality-flux-dev-lora

Make sure to play around with the strengths!

3 Workflow-Walkthrough

Prompt

After you downloaded and loaded all the models it is time to create the prompt. Be as detailed and descriptive as possible - Flux really likes that. Name the style, lighting, composition, action, look and feel of your image.

photography of a woman called a1mber standing in an italian village. she is wearing a grey turtleneck sweater, brunette hair, she is standing on an old bridge with views of a landscape in tuscany, candid photography, she is smiling. amateur photography with motion blur, iphone photo, hard sunlight, low resolution candid photo taken on a phone, grainy, the old rustic buildings are decorated in great detail, hard evening sunlight reflects from her skin

a1mber is the trigger word for my character lora!

ControlNet

A ControlNet is a tool that lets you guide AI image generation by using reference images or sketches to control specific aspects like pose, composition, or layout while still maintaining creative freedom in other elements.

Using it with this workflow is completely optional. Activate it using the fast bypasser in the beginning of the workflow.

I recommend using openpose or canny as inputs. I mostly use this online tool to pose my characters and then download the poses from there. Works really great most of the time, but don’t expect it to work every time. Sometimes you need to try out different seeds or increase the strength parameter in the “Apply Advanced ControlNet” node.

Sampler

  1. Set your scene dimensions here.

  2. The Detail Daemon Sampler is an insanely powerful tool to add detail to your generated images. I recommend checking out some tutorials for this tool alone, but make sure to play around with the detail_amount. I personally keep it between 0.10 - 0.60 since higher values can break the images but make sure to try it out yourself!

  3. batch_size: How many images you want to generate in one go.

  4. Noise seed for the image. Change it to get a completely new variation of your prompt. Set it to “randomize”

  5. These are the basic sampler settings and these work well for most images I generated. But feel free to play around with them and test out different sampler and scheduler combinations.

  6. Should be greater than 1. This value worked great for me but feel free to try out different values.

  7. This will save out a low res preview of your image. Deactivate this node using CTRL+B if you don’t need it.

Upscale

Here I am using an SDUltimate upscaler with Flux. This one will break down the image into smaller tiles and upscale them one after another creating more detail but also make it easier to process for your computer.

←I am using this combination of settings and I recommend you use them as a basis, but as always play around with them!

The most important value is the denoise value. This will change how much the upscaler is allowed to change your image. A value of 0.30 is a good starting point.

It will allow the sampler to change the image 30% giving it some freedom to change the image but also making sure that it stays consistent.

Post Processing

  1. Chromatic aberration is the visual distortion that occurs when a lens fails to focus all colors to the same point, causing colored fringes along boundaries between dark and light areas of an image.

  2. Image sharpening in phone photography is the process of enhancing edge contrast to make details appear more defined and crisp to compensate for limitations in small camera sensors.

  3. Bloom in phone photography is the halo-like glow that appears around bright light sources. Smudges on a phone lens can create or exaggerate bloom effects because fingerprints, oils, or dirt particles scatter incoming light in unpredictable directions.

  4. Film grain in phone photography is the deliberate addition of noise or texture to digital images to simulate the organic, slightly imperfect look of traditional film photographs.

  5. Image compression in phone photography is the process of reducing file sizes by selectively discarding visual data, often resulting in artifacts like blocky textures or loss of fine detail when overdone.

You can find the WORKFLOWS & EXAMPLE FILES here:

© 2025 Mickmumpitz

© 2025 Mickmumpitz

© 2025 Mickmumpitz