Generative Workflow Suite 1.4.1.1 – Consistent AI Image Generation & Editing

Overview

This suite of workflows is designed for professionals and creatives working in fashion, e-commerce, advertising, product design, and editorial content creation, offering a flexible set of tools to generate or edit consistent, high-quality AI images with minimal effort and maximum control.

Each workflow is structured to address a specific need:

• Some focus on automated generation from a single prompt file
• Others allow for manual creative control
• Some are built for multi-image compositions (e.g., model + product)
• Others support contextual editing of existing images, powered by Flux.1 Kontext

These workflows are modular, fast to deploy, and fully compatible with ComfyUI. They integrate GPT-powered prompt creation tools and can handle both single-shot creative tasks and large-scale batch automation.

✅ What's Included (Version 1.4.1)

Each workflow is designed to offer a different level of control, automation, and image conditioning. Whether you're working with a single reference or managing complex, multi-layered compositions, this release has you covered:

1.4.1_A – Automated batch generation from prompt file + 1 reference image
Fully automated image generation using a prompt text file (no line breaks) and a single reference image. Ideal for generating large batches using GPT-generated instructions.

1.4.1_B – Manual prompt input + 1 reference image
A simplified version of A. Allows manual prompt control with a single reference image. Great for custom testing and creative iterations.

1.4.1_C – Multi-image input + multi-prompt for complex compositions
Supports up to 6 input images (face, pose, outfit, product, etc.) and generates prompt-driven outputs using GPT logic. Perfect for campaigns needing strong talent + product consistency.

1.4.1_D – Multi-image input + single manual prompt
Gives you full creative freedom over a complex setup with manual control. Maintain reference conditioning while writing your own prompt.

1.4.1_E – Contextual editing using Flux.1 Kontext PRO & MAX
Uses Flux.1 API to edit a generated image contextually. Ideal for changing lighting, environment, styling, or repositioning without affecting subject consistency.

1.4.1_F – Gemini generation (1 image + 1 prompt) + Kontext PRO editing
First image is generated via Gemini (more consistent subject creation), then edited using Flux.1 Kontext for ambient, background or lighting shifts.

1.4.1_G – Gemini generation (multi-image input) + Kontext PRO editing
The most advanced flow: multi-image conditioning (e.g., face + outfit + product) generates a highly accurate base image via Gemini, then passes to Kontext for scene transformation or enhancement.

1.4.1_H – Canvas-based image merging + Kontext PRO refinement
A hybrid workflow that lets you compose two or more images (e.g. model + garment) on canvas using Canvas as Image, then pass the result to Flux.1 Kontext PRO. Perfect for generating realistic “wearing” effects, scene composites, and smart product applications.

In Progress: Upcoming Simplified Versions

I’m currently working on streamlined workflows tailored to more specific use cases, including:

• Outfit try-on for fashion brands and stylists
• Product scene remixes (e.g. swapping items in hand, background localization)
• Talent consistency testing for character design pipelines
• Editorial layout support with advanced typography and layout editing via Kontext MAX

Each future module will follow the same logic of minimal configuration + maximum quality, supporting both automation and manual fine-tuning.

Workflow 1.4.1_A – AI Image Automation from Single Reference

Introduction

This workflow is designed for professionals in e-commerce, fashion, editorial, or product catalog fields who need to generate a high volume of consistent AI images starting from a single reference photo. It’s ideal for scalable visual production where poses, outfits, or backgrounds change, but the identity and consistency of the character must remain intact.

Thanks to integration with custom GPTs, the entire prompt-writing process is fully automated, eliminating the need for manual input.

What This Workflow Does

Workflow 1.4.1_A allows you to generate AI images from:

• A single reference image (e.g., a portrait or full-body shot of a model)
• A .txt prompt file, automatically generated using one of the following GPTs:
  • Character Prompt Catalogue
  • Character + Product
  • Pose Only
  • Pose + Outfit

The system reads the .txt file line by line—each line containing a single prompt—and uses it to generate a batch of consistent images based on the loaded reference.

Key Features

• Input: one reference image + one .txt file (no line breaks, one prompt per line)
• Output: a sequence of high-quality, consistent images
• Fully automated: no manual interaction required during generation
• Perfectly integrated with GPTs for structured, professional scene descriptions

Included Files

• ✅ Preconfigured .json workflow file (compatible with ComfyUI v0.3.27+)
• ✅ Example reference image: Mahina.png
• ✅ Sample prompt file: Character_Mahina.txt
• ✅ Direct links to custom GPTs for prompt generation

How It Works (Overview)

1. Load a reference image (portrait or full-body)
2. Generate a .txt file with one prompt per line (no line breaks)
3. Save the file in /ComfyUI/input/
4. Launch the workflow: it will automatically process the prompts and output a sequence of consistent images

Required Modules and Resources

🔧 Required ComfyUI Nodes

• Text Load Line From File
  ↳ From was-node-suite-comfyui
• easy showAnything
  ↳ From ComfyUI-easy-use
• SetNode, GetNode, ShowText, Number Counter, PreviewImage, SaveImage
  ↳ Default nodes included in ComfyUI
• DualCLIPLoader, UNETLoader, VAELoader
  ↳ Requires additional models (see below)

📦 Models to Download

Generation Model

• flux1-dev.safetensors
  ↳ Main model for consistent generation
  ↳ Location: /ComfyUI/models/unet/
  ↳ Download from: Hugging Face or official Flux1 repo (depending on license availability)

Tokenizers

• clip_l.safetensors
  ↳ Location: /ComfyUI/models/clip/
• t5xxl_fp16.safetensors
  ↳ Location: /ComfyUI/models/conditioner/ (or equivalent folder)

Practical Example

In the download package, you'll find Mahina.png and a prompt file Character_Mahina.txt. When you run the workflow, it will generate several consistent outputs based on that image—perfect for catalog content, character tests, or scalable production use.

Workflow 1.4.1_B – Manual Prompt Input from Single Reference

Introduction

This workflow is ideal for users who prefer full manual control over each prompt, enabling the creation of individual AI images based on a single reference photo and a custom-written text description. It’s perfect for editorial, fashion, or product use cases where creative direction changes from image to image and automated batch generation isn’t required.

Compared to version 1.4.1_A, this setup is streamlined and more flexible for exploratory creative sessions or one-off image productions.

What This Workflow Does

Workflow 1.4.1_B allows you to generate a single AI image at a time using:

• A single reference image (portrait or full-body)
• A manually written prompt, input directly into the workflow interface

Unlike version A, this workflow does not read from a .txt batch file. Instead, it lets you test and iterate on prompt language in real-time, making it ideal for fine-tuning poses, outfits, lighting, or compositions.

Key Features

• Input: one reference image + one manual prompt
• Output: one high-quality image per run
• Real-time prompt control: edit the prompt directly inside ComfyUI
• Designed for precision: ideal for visual exploration and creative direction

Included Files

• ✅ Preconfigured .json workflow file
• ✅ Example reference image (e.g., Mahina.png)
• ✅ Prompt examples (can be copy-pasted manually)
• ✅ Compatible with GPT-generated prompts if desired

How It Works (Overview)

1. Load your reference image (e.g., a model photo)
2. Paste or type a custom prompt directly into the Set Prompt node
3. Run the workflow and generate a single image
4. Modify the prompt and repeat to produce variants on demand

Required Modules and Resources

🔧 Required ComfyUI Nodes

• SetNode, GetNode, ShowText, PreviewImage, SaveImage
  ↳ Default nodes included in ComfyUI
• DualCLIPLoader, UNETLoader, VAELoader
  ↳ Requires specific models (see below)

📦 Models to Download

Generation Model

• flux1-dev.safetensors
  ↳ Main model for consistent generation
  ↳ Location: /ComfyUI/models/unet/
  ↳ Download from: Hugging Face or official Flux1 repo

Tokenizers

• clip_l.safetensors
  ↳ Location: /ComfyUI/models/clip/
• t5xxl_fp16.safetensors
  ↳ Location: /ComfyUI/models/conditioner/ (or equivalent folder)

Practical Example

You can use the same reference image (e.g., Mahina.png) as in Workflow A, but instead of relying on a .txt batch of prompts, you can manually input a prompt like:

"Rear full-body studio portrait of a tall, elegant woman in a short red satin kimono. Soft lighting on a white seamless background. Neutral pose, calm mood, fashion editorial style."

This setup gives you maximum flexibility while still maintaining character consistency from your image reference.

Workflow 1.4.1_C – Multi-Prompt + Multi-Image Composition

Introduction

This workflow is designed for advanced image generation tasks where the goal is to combine a consistent character (talent) with multiple reference images, such as product visuals, outfits, or props. It is particularly useful for projects in advertising, fashion, cosmetics, grooming, or tech accessories, where multiple visual elements need to be composed together in a coherent and realistic scene.

Use cases include:

• A model interacting with a product (e.g. perfume, trimmer, smartphone, hairdryer)
• Outfit try-ons across different poses or environments
• Talent consistency matched to multiple objects or backgrounds

What This Workflow Does

Workflow 1.4.1_C enables the automated generation of consistent images by combining:

• Multiple reference images (up to 6) — e.g. one of a model, others of products or garments
• Multiple prompts, one for each image or scene setup

The prompts follow the same structure used in workflow 1.4.1_A, and are typically generated using GPTs from the Character Prompt Catalogue series. This ensures each image maintains identity and visual harmony while adapting poses, interactions, and product placements dynamically.

Key Features

• Input:
  • Up to 6 images (e.g., model + 5 products or outfit views)
  • A .txt file with 1 prompt per line, synced to each image
• Output: a sequence of high-quality, character-consistent image compositions
• Optimized for commercial use: perfect for ad campaigns, product listings, lookbooks
• Automated workflow: batch image generation driven by descriptive GPT-written prompts

Included Files

• ✅ Preconfigured .json workflow file
• ✅ Folder structure example for multi-image input
• ✅ Prompt .txt file template for synchronized descriptions
• ✅ Example image set (model + 2–3 product shots)

How It Works (Overview)

1. Load up to 6 input images (one per visual element: model, outfit, products, etc.)
2. Create a prompt .txt file (no line breaks, one prompt per row, matching the image order)
3. Save the prompt file in /ComfyUI/input/
4. Run the workflow to generate each composition based on the corresponding prompt-image pair

Each image produced is a unique blend of consistent subject features (e.g., face, body, lighting) and visual context (pose, background, object interaction).

Required Modules and Resources

🔧 Required ComfyUI Nodes

• Text Load Line From File
  ↳ From was-node-suite-comfyui
• Any Switch (rgthree)
  ↳ Allows multiple image inputs to be routed dynamically
  ↳ GitHub – rgthree tools
• SetNode, GetNode, ShowText, PreviewImage, SaveImage
  ↳ Core nodes (default in ComfyUI)
• DualCLIPLoader, UNETLoader, VAELoader
  ↳ Requires additional models (see below)

📦 Models to Download

Generation Model

• flux1-dev.safetensors
  ↳ For consistent talent generation
  ↳ Location: /ComfyUI/models/unet/
  ↳ Download from: Hugging Face or official Flux1 repo

Tokenizers

• clip_l.safetensors
  ↳ Location: /ComfyUI/models/clip/
• t5xxl_fp16.safetensors
  ↳ Location: /ComfyUI/models/conditioner/ (or equivalent)

Practical Example

A typical use case could include:

• image_01: the model reference
• image_02–image_06: product shots (e.g., perfume bottle, trimmer, smartphone)
• Prompt file:

  Elegant woman holding a luxury perfume bottle, soft backlight in a white studio
  Fashion model with smartphone in hand, urban lifestyle setting
  ...
  

Each output image will match a different combination, enabling fast creation of talent + product shots with consistent styling and realism — suitable for both editorial spreads and product pages.

Workflow 1.4.1_C – Multi-Prompt + Multi-Image Composition

Introduction

This workflow is designed for advanced image generation tasks where the goal is to combine a consistent character (talent) with multiple reference images, such as product visuals, outfits, or props. It is particularly useful for projects in advertising, fashion, cosmetics, grooming, or tech accessories, where multiple visual elements need to be composed together in a coherent and realistic scene.

Use cases include:

• A model interacting with a product (e.g. perfume, trimmer, smartphone, hairdryer)
• Outfit try-ons across different poses or environments
• Talent consistency matched to multiple objects or backgrounds

What This Workflow Does

Workflow 1.4.1_C enables the automated generation of consistent images by combining:

• Multiple reference images (up to 6) — e.g. one of a model, others of products or garments
• Multiple prompts, one for each image or scene setup

The prompts follow the same structure used in workflow 1.4.1_A, and are typically generated using GPTs from the Character Prompt Catalogue series. This ensures each image maintains identity and visual harmony while adapting poses, interactions, and product placements dynamically.

Key Features

• Input:
  • Up to 6 images (e.g., model + 5 products or outfit views)
  • A .txt file with 1 prompt per line, synced to each image
• Output: a sequence of high-quality, character-consistent image compositions
• Optimized for commercial use: perfect for ad campaigns, product listings, lookbooks
• Automated workflow: batch image generation driven by descriptive GPT-written prompts

Included Files

• ✅ Preconfigured .json workflow file
• ✅ Folder structure example for multi-image input
• ✅ Prompt .txt file template for synchronized descriptions
• ✅ Example image set (model + 2–3 product shots)

How It Works (Overview)

1. Load up to 6 input images (one per visual element: model, outfit, products, etc.)
2. Create a prompt .txt file (no line breaks, one prompt per row, matching the image order)
3. Save the prompt file in /ComfyUI/input/
4. Run the workflow to generate each composition based on the corresponding prompt-image pair

Each image produced is a unique blend of consistent subject features (e.g., face, body, lighting) and visual context (pose, background, object interaction).

Required Modules and Resources

🔧 Required ComfyUI Nodes

• Text Load Line From File
  ↳ From was-node-suite-comfyui
• Any Switch (rgthree)
  ↳ Allows multiple image inputs to be routed dynamically
  ↳ GitHub – rgthree tools
• SetNode, GetNode, ShowText, PreviewImage, SaveImage
  ↳ Core nodes (default in ComfyUI)
• DualCLIPLoader, UNETLoader, VAELoader
  ↳ Requires additional models (see below)

📦 Models to Download

Generation Model

• flux1-dev.safetensors
  ↳ For consistent talent generation
  ↳ Location: /ComfyUI/models/unet/
  ↳ Download from: Hugging Face or official Flux1 repo

Tokenizers

• clip_l.safetensors
  ↳ Location: /ComfyUI/models/clip/
• t5xxl_fp16.safetensors
  ↳ Location: /ComfyUI/models/conditioner/ (or equivalent)

Practical Example

A typical use case could include:

• image_01: the model reference
• image_02–image_06: product shots (e.g., perfume bottle, trimmer, smartphone)
• Prompt file:

  css

  CopiaModificaElegant woman holding a luxury perfume bottle, soft backlight in a white studio Fashion model with smartphone in hand, urban lifestyle setting ...

Each output image will match a different combination, enabling fast creation of talent + product shots with consistent styling and realism — suitable for both editorial spreads and product pages.

Workflow 1.4.1_D – Multi-Image Input with Manual Prompt

Introduction

This workflow is ideal for scenarios where you want to compose an AI image using multiple visual references—such as a model and one or more products—while maintaining full manual control over the text prompt. It's especially useful in creative or commercial contexts where the setup is custom for each image, but still requires multiple inputs to condition the result.

Use cases include:

• Advertising shots combining a model with a perfume, trimmer, or electronic product
• Outfit try-ons where both the pose and outfit are provided as images
• Controlled scene composition using visual guidance plus precise prompt writing

What This Workflow Does

Workflow 1.4.1_D enables the creation of a single AI image by combining:

• Multiple input images (up to 6), e.g.:
  • A model or talent reference
  • One or more product images
  • An outfit or pose reference
• One manually written prompt to define the desired scene

This approach is a manual counterpart to workflow 1.4.1_C, offering more flexibility for single-scene fine-tuning, storytelling, or editorial creation where automated batch processing isn't needed.

Key Features

• Input:
  • Up to 6 images (e.g., model + products or outfit views)
  • One manual text prompt
• Output: a single image composition based on all visual and textual cues
• Ideal for controlled creative direction and fine-tuning output
• Consistent identity & design logic, especially for model + product scenarios

Included Files

• ✅ .json workflow file for ComfyUI
• ✅ Example multi-image setup
• ✅ Prompt writing guide and sample text
• ✅ GPT-compatible structure for copy-paste prompt generation

How It Works (Overview)

1. Load up to 6 reference images (model, outfit, product, pose...)
2. Paste your custom prompt into the Set Prompt node
3. Launch the workflow to produce a single coherent AI image
4. Adjust the image set or prompt as needed to create new variants

Required Modules and Resources

🔧 Required ComfyUI Nodes

• Any Switch (rgthree)
  ↳ Allows routing of multiple image inputs
  ↳ GitHub – rgthree tools
• SetNode, GetNode, ShowText, PreviewImage, SaveImage
  ↳ Default ComfyUI nodes
• DualCLIPLoader, UNETLoader, VAELoader
  ↳ Requires additional models (see below)

📦 Models to Download

Generation Model

• flux1-dev.safetensors
  ↳ High-consistency model for character-based workflows
  ↳ Location: /ComfyUI/models/unet/
  ↳ Available on Hugging Face or Flux1 GitHub

Tokenizers

• clip_l.safetensors
  ↳ Location: /ComfyUI/models/clip/
• t5xxl_fp16.safetensors
  ↳ Location: /ComfyUI/models/conditioner/

Practical Example

You might load:

• image_01: model reference (Mahina)
• image_02: red satin kimono
• image_03: perfume bottle
• Prompt:

  “Full-body editorial shot of a tall, elegant woman in a red satin robe holding a perfume bottle. Clean studio background, soft light, luxury magazine style.”

This allows you to merge reference elements visually and control the result with professional-level prompt crafting.

Workflow 1.4.1_E – Contextual Editing with Flux.1 Kontext PRO & MAX

Introduction

Workflow 1.4.1_E introduces the use of Flux.1 Kontext, a cutting-edge system for image-to-image editing in the latent space. Unlike other generation workflows, this pipeline allows you to start from an existing image and modify it with precision—changing elements, styles, accessories, or even entire compositions—while preserving the original structure, character identity, pose, lighting and camera perspective.

This is ideal for:

• Fashion campaigns with outfit variations
• Product replacement or enhancement (e.g. replacing a perfume bottle or glasses)
• Facial expression, pose, background or accessory changes
• Advertising creatives where one base shot is reused in many ways

What This Workflow Does

This workflow has two editing branches:

🔹 Kontext PRO

• Lower cost ($0.04 per image)
• Optimized for structural editing, style transfer, background change, and precise object-level modifications
• Ideal for: makeup edits, outfit variations, background replacement, hair color/style changes

🔸 Kontext MAX

• Higher cost ($0.08 per image)
• Optimized for prompt precision, typography, text editing, and faster convergence
• Ideal for: editorial layout editing, logo/text replacement, highly detailed creative scenes

Both branches accept:

• A source image as the base
• A prompt that tells Flux.1 what to change, add, remove, or preserve

Key Features

• Input:
  • One reference image (photographic, fashion, product or character)
  • One natural language instruction (prompt)
• Output:
  • A modified image preserving original quality and structure
• Two modes:
  • Kontext PRO for accuracy and consistency
  • Kontext MAX for speed and prompt sensitivity
• Advanced latent editing: no need to regenerate entire scenes from scratch

Included Files

• ✅ .json workflow with preloaded PRO and MAX editing branches
• ✅ Prompt templates and usage examples
• ✅ Sample input images (optional)

Example Prompts

Here are real examples of effective prompts tailored for Flux.1 Kontext workflows:

➤ Fashion Example

Input: Full-body studio photo of a model in red satin kimono
Prompt:

"Replace the robe with a long black leather trench coat. Keep lighting, pose, and identity unchanged."

➤ Product Swap

Input: Woman holding a perfume bottle
Prompt:

"Replace the perfume bottle with a smartphone. Keep her hands and body position consistent."

➤ Makeup / Facial Edit

Input: Portrait of a female character
Prompt:

"Add smokey eye makeup and matte red lipstick. Do not change her hairstyle or expression."

➤ Background Replacement

Input: Model walking in studio
Prompt:

"Place her on a sunny street with modern glass buildings. Maintain same camera angle and lighting on her body."

Required Modules and Resources

🔧 Required ComfyUI Nodes

• Flux.1 API Node – Kontext PRO
• Flux.1 API Node – Kontext MAX
  ↳ ComfyUI Flux.1 Kontext Pro Image API Node Official Example - ComfyUI
• SetNode, GetNode, ShowText, PreviewImage, SaveImage
  ↳ Default ComfyUI modules

📦 External Configuration

API Key Required

To use Kontext nodes, you'll need to:

• Register at: https://runwayml.com or the official Flux provider
• Insert your API key into the node or set as environment variable
• Select mode: kontext_pro or kontext_max

🆕 New in this Version: Multi Image Flux.1: Kontext PRO

We’ve added a new Multi Image Flux.1: Kontext PRO module (0.04 USD per call), which allows you to inject two supporting images alongside your main image. This enables advanced compositional merging, style transfer, and hybrid scene creation with high fidelity and control.

📌 You can define:

• Main image: the primary subject to edit
• Image 1 + Image 2: additional images whose elements (e.g. background, lighting, objects) will be used to condition the final result

🧩 Required Node:

To use the multi-image module, you need to install the following custom node:

🔗 https://github.com/grinlau18/ComfyUI_XISER_Nodes

This node enables structured multi-image input support for Flux.1 Kontext in ComfyUI, as described in the official API integration guide.

⚙️ Use Cases:

• Change the lighting and mood of a model photo using a separate ambient reference
• Combine an interior shot with an outdoor background in a consistent visual tone
• Place a product (perfume, phone, appliance) into a new composition with the right shadows and reflections
• Match a generated fashion photo to a real-world location or branded environment

✍️ Prompt Example (customized):

Combine the main image with the ambiance and background elements of Image 1. Preserve the pose, lighting direction, and facial structure from the main image. Use the textures and warm tone of Image 2 to influence the environment. The final result should look like a high-end magazine editorial photo.

🧠 Tips:

• Use high-quality and stylistically consistent source images for better results
• The main subject will retain its identity, but context (environment, lighting, mood) will be enhanced or transformed based on input references
• Kontext PRO is best for structure and scene composition, while Kontext MAX is ideal for fine-tuning layout, typography, and localized prompt control

Workflow 1.4.1_F – Gemini + Kontext PRO (Single Image & Prompt)

Introduction

This hybrid workflow combines the precision and character consistency of Google Gemini with the editing and compositional power of Flux.1 Kontext PRO. It is ideal for scenarios where you want to generate a clean, high-quality base image (e.g., a model or a car) and then edit that image to modify lighting, environment, background, or product details.

Use cases include:

• Creating a fashion talent with Gemini, then placing them in different editorial environments
• Generating a product (e.g. a white car, smartphone, perfume bottle) with Gemini, then customizing it with Kontext

What This Workflow Does

Workflow 1.4.1_F performs in two sequential steps:

1. Gemini Image Generation
   • Uses a single prompt to generate a high-quality base image
   • Ensures identity consistency, sharpness, and structured visuals
2. Kontext PRO Editing
   • Modifies the Gemini image based on a natural language prompt
   • Allows changes to lighting, setting, background, styling, or minor elements

Key Features

• Input:
  • One manual prompt
  • (Internally) uses the Gemini generation engine
• Editing:
  • Kontext PRO applied to the generated image
• Output:
  • One coherent, edited image with consistent structure
• Perfect for creative variation: start with one subject, evolve into many settings or moods

Example Use Cases

Use CaseGemini OutputKontext Prompt ExampleFashion talent generationWoman in silk dress on white backdrop"Place her in a nighttime city street, neon lights behind"Car rendering and customizationWhite Porsche GT3, studio lit"Replace the livery with red Marlboro style. Trackside shot"Tech product editorialBlack smartphone, neutral studio"Add soft backlight, place on marble table in daylight"

Required Modules and Resources

• ✅ Gemini API Node (single prompt) – image generation
• ✅ Flux.1 Kontext PRO Node – editing layer
• ✅ Standard SetNode, ShowText, PreviewImage, SaveImage
• 🔑 Gemini API key (VPN may be required depending on region)

Workflow 1.4.1_G – Gemini + Kontext PRO (Multi-Image + Prompt)

Introduction

Workflow 1.4.1_G extends the hybrid Gemini-Kontext logic into multi-image generation and editing. It allows you to use multiple visual elements (e.g. character, outfit, product) and compose them into a single image via Gemini, followed by precise contextual editing via Flux.1 Kontext PRO.

This is the go-to solution for complex AI scene building, particularly when working with:

• Characters + outfits + accessories
• Vehicles with configurable livery
• Beauty products, devices, or furniture with custom environments

What This Workflow Does

1. Gemini Multi-Image Generation
   • Combines multiple images (up to 6) and a single prompt
   • Produces a consistent, coherent image of a character or product
2. Kontext PRO Editing
   • Modifies the resulting image with text prompts
   • Useful for location change, light variation, seasonal effects, and more

Key Features

• Input:
  • Up to 6 image references (pose, outfit, product...)
  • One structured prompt
• Editing:
  • Kontext PRO as final step
• Output:
  • Stylized, customized AI image with high fidelity and character consistency
• Best for storytelling and multi-scene campaigns

Example Use Cases

ScenarioMulti-InputKontext Prompt ExampleTalent + Outfit + ProductModel + red dress + perfume image"Place her outdoors at sunset, soft flare behind"Car with modifiable liveryWhite base car + 2 decal references"Replace graphics with Alitalia livery, street background"Gadget launch visualsPhone + case + hand model"Scene at a café table, morning light, cozy interior"

Required Modules and Resources

• ✅ Gemini API Node (multi-image)
• ✅ Flux.1 Kontext PRO Node
• ✅ Any Switch, Image Merge, or equivalent multi-input setup
• 🔑 Gemini API (US VPN required in restricted regions)
• 🔑 Flux.1 API Key

Workflow 1.4.1_H – Canvas-Based Composition + Kontext PRO Editing

Introduction

Workflow 1.4.1_H introduces a powerful hybrid method that merges visual canvas composition with Flux.1 Kontext PRO editing, allowing you to merge two or more images inside ComfyUI, reposition them freely, and then transform the result using contextual prompts.

It is especially effective for virtual try-on scenarios, where a model image and a garment image are composed into a single canvas and refined to create a photorealistic "worn" effect.

What This Workflow Does

This workflow performs in two major stages:

🧱 1. Canvas Composition via Multi-Image Merge

• Uses Canvas as Image node from the ComfyUI_XISER_Nodes extension
• Allows positioning of multiple visual elements (e.g., model, outfit, accessory) on a shared canvas
• Automatically aligns elements using a white-background mask when available (e.g., cropped product image)
• Output: a single composited image ready for editing

🎯 2. Flux.1 Kontext PRO Editing

• The composited canvas is passed to Kontext PRO
• A natural language prompt is used to apply realistic lighting, adjust fabric behavior, remove overlaps, and enhance coherence
• Maintains identity, lighting, and camera angle while adding depth and realism

Key Features

Input:

• 1 reference canvas composed of:
  • A model (main subject)
  • A product (e.g. dress, glasses, tech device)
  • Optional overlays (e.g. masks, patterns)

Editing:

• Kontext PRO interprets the scene and applies advanced prompt-driven editing

Output:

• Realistic visual of a product being "worn" or physically integrated onto the subject
• High-end editorial or e-commerce result without manual retouching

Example Prompt

"Preserve the woman's face, frontal pose, and full-body composition. Keep the neutral studio lighting. Change the red and white striped crochet dress so it looks physically worn: adjust for realistic fabric fall, visible shoulder seams, natural tension around waist and hips, and accurate shadows cast by the dress. Remove any signs of overlay or visual detachment between the garment and body."

Example Use Case (Image Above)

• Input images:
  • A model with neutral lighting
  • A product photo (cropped dress on white background)
• Canvas Composition: both are merged and aligned in ComfyUI
• Flux.1 Kontext Prompt: transforms the flat overlay into a realistic, naturally fitting dress

Required Modules and Resources

✅ Canvas as Image node from ComfyUI_XISER_Nodes
✅ Flux.1 Kontext PRO node from RunwayML/Flux.1 integration
✅ WAS Node Suite for image utilities and formatting
✅ (Optional) JoyCaption2 Advance for post-upscale quality
🔑 Flux.1 API key (required for contextual editing)
🌍 VPN set to US region (if needed for API accessibility)

🤖 Need help writing prompts?

Use the Kontext Prompt Assistant:
🧠 https://chatgpt.com/g/g-683e7dc5e6008191b0fa539758152787-kontext-prompt-assistant

This GPT will help you write smart, context-aware instructions for every image you want to edit.