Ever wanted to merge the perfect outfit with the ideal pose, or showcase different products all in a single, seamless scene? Whether you’re a designer looking to preview fashion combinations, a marketer building composite product showcases, or an artist bringing creative concepts to life, this workflow gives you the power to experiment freely and see instant, consistent results.

What is Qwen Image Edit 2509? Why is it better?

Qwen Image Edit 2509 lets you combine multiple images into a single scene using text prompts. Reference different images by number in your prompt-"image 1," "image 2," "image 3"-and the model places them together.

Released in September 2025 (the "2509" refers to the version and release date), this update handles complex multi-image edits better than the original Qwen. Faster processing, better alignment between elements, and more realistic results when combining different inputs.

Useful for product mockups, fashion previews, or any project where you need to merge different elements into one cohesive image.

Why settle for one-at-a-time edits, when your next masterpiece could start with everything you need, all at once?

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 4 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set the Models Set the required models as seen in the image.
2. Load the Input Image Upload your input image. You can you either 1 up to 3 images only. It works even 1 input image only and just bypassed the other node.
3. Write the Prompt Write a detailed prompt. The should designate an image < number> in the prompt. You describe whatever you want and whatever action will be.
4. Check the Sampling Settings Check the sampling settings. If you want higher quality of output you can the full model but it needs a higher machine than Ultra. You can use the fp8 model instead.
5. Check the Output Image

Insights

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5510 members

Discord

Flux Krea Dev is built to create photorealistic images without the usual AI giveaways-no plastic skin, no overly smooth textures, no uncanny valley weirdness.

This is a collaboration between Black Forest Labs and Krea AI. It's a 12-billion parameter model that handles realistic skin tones, natural lighting, and follows prompts accurately. Works especially well for portraits and character shots.

Faster than standard Flux Dev and produces more varied, lifelike results.

What is Flux Krea Dev

Flux Krea Dev is designed to generate photorealistic images while avoiding common AI artifacts like plastic-looking skin or overprocessed textures. The 12-billion parameter model maintains realistic detail, handles nuanced lighting well, and produces natural skin tones.

It supports fine-tuning, works with existing Flux workflows, and generates images quickly. Good for portraits, character work, and any project where you need genuinely realistic human features.

Comparison of Flux Krea Dev with Standard model

Flux Krea Dev is a major upgrade from the original Flux Dev, designed to create genuinely lifelike images. Where Flux Dev often produces flat, repetitive, or generic results, Flux Krea Dev excels in capturing photorealistic detail, prompt accuracy, and expressive variety—making it the best open-source choice for anyone seeking high-quality, realistic AI-generated art and portraits.

Get ready for a hands-on experience designed to make your art leap off the screen and make people ask, “Is that really AI?”

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 4 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set the Models Set the models as seen on the image.
2. Write a Prompt Write a prompt that describes a realistic description. It is better to depict a character or portrait inthe prompt.
3. Check the Sampling Set the settings as seen on the image. Don't change the scheduler or sampler as it may result to bad quality.
4. Check the Generated Image

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

If you're having issues with workflow and visit us here at Discord #Help Desk or you may opt to email us at [email protected]

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5449 members

Discord

Flux is one of the most popular AI image models in ComfyUI right now. It handles text-to-image generation well, follows prompts accurately, and works across different art styles.

Here are the 5 Flux workflows people are using the most. Each one does something different-from filling in parts of images to training custom models to keeping characters looking consistent across scenes.

Pick whichever matches what you're trying to do.

Inpainting with Reference using Flux Fill and Flux Redux

What it's great for:

• Seamlessly inpaints, outpaints, and fills missing or extended areas with natural blending and high visual consistency.
• Accurately follows user text prompts for content replacement, restyling, and creative modifications
• Combines styles or images to generate unique variations while preserving important details.
•  Supports adjustments in aspect ratio, guidance scale, resolution, and blending intensity for tailored results.

Inpainting using reference images is ideal for digital artists, photographers, designers, and content creators who want to restore, enhance, or modify images seamlessly. It benefits anyone needing to remove objects, repair photos, or add new elements that match the original style-making it useful for both professionals and hobbyists.

The Inpainting Revolution: How Reference Images with Flux Fill and Flux Redux Are Changing the Game

Ever needed to add something to an image that wasn’t there before? That’s where Flux Fill and Flux Redux come in – they’re changing the game for image editing by making inpainting (filling in parts of images) look natural and professional. By using models such as Flux Fill and Flux Redux,

ThinkDiffusionPhu Ngo

Train Flux Models using Flux

What it's great for:

• Train Flux models efficiently on machine with low VRAM.
• Node-based interface to manage the entire training process within the same environment as your image generation workflows.
• Easily train on small datasets (10–30 images), mix multiple datasets, and apply data augmentation or custom captions for diverse results.
• Fine-tune training with adjustable parameters like batch size, learning rate, optimizer type.
• Produce detailed and consistent models suitable for portraits, objects, styles, and more.

Training Flux models with ComfyUI is great for digital artists, designers, developers, and content creators who want to create custom AI image styles or assets. It’s accessible even for beginners and hobbyists, thanks to its user-friendly interface and low hardware requirements, making advanced AI image generation possible for anyone interested in creative or professional projects.

Building Better Models: Flux LoRAs in ComfyUI

What if you could make every image you generate, conform to a certain style or person? This is exactly what using a LoRA model with Flux AI in ComfyUI does. In this guide, we’ll explore how Flux can help you build stronger, more efficient models with ease. Whether you’re new

ThinkDiffusionPhu Ngo

Image2Image using Flux Controlnet

What it's great for:

• Integrates edge detection (Canny, HED) and depth maps as information to guide image transformation.
• All models are trained and optimized for 1024x1024 resolution, enabling the generation of detailed, high-quality images suitable for professional and creative use.

Image2Image Flux with ControlNet is ideal for digital artists, designers, animators, and content creators who want precise control over image transformations. It benefits anyone looking to guide AI-generated images with edge, pose, or depth maps-making it useful for creative projects, marketing visuals, game assets, or personal artwork.

Precision in Flux Art: Harnessing the Power of ControlNet

Flux lets you create impressive images from text prompts. ControlNet is a significant tool for controlling the precise composition of AI images.

ThinkDiffusionPhu Ngo

Consistent Character Creating using Flux

What it's great for:

• Generates multiple character poses and expressions from a single reference image.
• Maintains strong identity consistency across all outputs (face, clothing, colors).
• Supports multi-angle and multi-scene character generation for comics, games, and animation.
• Allows easy style, background, and attribute customization with prompts.

Artists, animators, game developers, marketers, and hobbyists can all benefit from Flux’s consistent character workflow. It’s perfect for anyone who needs uniform character designs across multiple scenes, poses, or styles-making it useful for comics, animation, games, branding, and personal creative projects.

Utilizing Flux in ComfyUI for Consistent Character Creation

Creating characters that look the same every time is crucial. This guide will show you, to maintain consistency by using the workflow in ComfyUI.

ThinkDiffusionPhu Ngo

Intro to Flux

What it's great for:

• Ensures consistent character appearance across multiple images and poses.
• Works with both reference images and text prompts for flexible input.
• Offers modular, user-friendly workflows for generation, upscaling, and detailing.
• Supports various art styles and creative applications like comics, games, and animation.
• Provides advanced controls for fine-tuning and precise customization.

Flux in ComfyUI is great for digital artists, designers, marketers, content creators, and hobbyists. Anyone who needs quick, high-quality, and consistent AI-generated images for creative projects, branding, or personal use can benefit from using it.

Introduction to Flux - Quick Guide

Flux has bursted on the scene as the defacto AI Art model. It’s here and easy to use on ThinkDiffusion, let’s dive in and show you how it works!

ThinkDiffusionPhu Ngo

If your computer's struggling or installation is giving you headaches, try these workflows in your browser with ThinkDiffusion. We provide the GPU power so you can focus on creating.

Enjoy experimenting with these workflows! And remember - every pro started as a beginner once.

If you enjoy ComfyUI and you want to test out HyperSD in ComfyUI and Blender in real-time, then feel free to check out this Real-Time Creativity: Leveraging Hyper SD and Blender with ComfyUI. And have fun out there with your videos!

Type a scene description, get a video. That's Wan2.2.

This is an open-source text-to-video model that uses a Mixture-of-Experts system to create realistic motion and accurate visuals. It handles 720p videos with smoother animation and fewer artifacts than version 2.1, and it runs on standard GPUs without needing a server farm.

Describe a fantasy world, a dramatic scene, or something everyday—Wan2.2 turns it into video. Works well for storyboarding, concept testing, or just seeing your ideas move.

What we'll cover

1. What Wan2.2 is and how it's better than 2.1
2. Getting the workflow running on ThinkDiffusion
3. Installing the 6 models you need
4. Walking through the workflow settings
5. Real video examples across different styles
6. Common issues and fixes

The Wan2.2 Release

Wan2.2 is a next-generation, open-source text-to-video model featuring a Mixture-of-Experts (MoE) system for dramatically more realistic motion, prompt accuracy, and cinema-quality visuals compared to Wan2.1. With much larger training data and smart MoE design, it delivers fluid, artifact-free 720p videos quickly and efficiently, even on standard GPUs. Artists, animators, filmmakers, and creators at any level will find Wan2.2 superior for its greater detail, smoother animation, and enhanced creative control—making it the go-to tool for high-quality, prompt-driven video generation.

Whether you’re experimenting with fantasy worlds, dramatic scenes, or lifelike moments that pulse with real atmosphere, prepare to be amazed—because with Wan2.2, the magic of cinematic video is just a sentence away waiting to be brought to life by your imagination alone.

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 6 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set the Models Set the models as seen on the image. Enabled the low mem load if you experienced the out of memory.
2. Write a Prompt Write a detailed prompt. Wan2.2 model is good in adherence of prompt. Set the size to 480p only. Wan2.2 is compatible with 720 and 1080 resolution and higher frames but you need a higher machine for that.
3. Check Sampling Set the sampling setting as seen on the image. Since it uses a lightning x2v lora, the inference steps should be at 4 only. Otherwise, it will result to an error.
4. Check the Video

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory (Allocation on Device error): Use smaller expansion factors, lower resolution, or less frames. Optionally try higher VRAM machines such as ULTRA or newly available NITRO (beta feature). Note that even with more VRAM, you may still encounter out-of-memory issues with Wan2.2.
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

If you're having issues with workflow and visit us here at Discord #Help Desk or you may opt to email us at [email protected]

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5471 members

Discord

Change backgrounds. Swap objects. Add stuff. Remove stuff. Adjust styles. All through simple text prompts instead of wrestling with complicated tools.

Qwen is Alibaba's image editing model, built on their 20B-parameter foundation. It handles object manipulation, style transfers, and even text editing inside images. The results are surprisingly realistic, and it keeps context better than you'd expect.

Useful if you're creating content, designing stuff, running social media, or just want to edit images without learning Photoshop.

What we'll cover

1. What Qwen Image Edit actually is and what makes it different
2. Getting the workflow running on ThinkDiffusion
3. Installing the models and custom nodes you need
4. Walking through the workflow step-by-step
5. Real examples of what it can do
6. Common issues and how to fix them

What is Qwen Image Edit?

Qwen Image Edit is a model developed by Alibaba's Qwen team, built upon their robust 20B-parameter Qwen-Image foundation. This model brings precise object manipulation, accurate style and background transfer, and dual-language text editing directly within images. It does a solid job with realism and keeping details intact, even when you're asking it to do tricky edits.

Ideal for content creators, designers, marketers, social media teams, localization experts, e-commerce businesses, and anyone seeking intuitive, professional-grade image editing through the power of natural language.

So go ahead: give your edits a voice, and see just how far your words can take your next photo adventure!

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 3 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load an Input Image Load an input image. Any minimal resolution will do as long it has a high quality image.
2. Set the Models Set the models as seen on the image. If you have good hardware you can use the full model but needs to be manual download.
3. Write the Prompt Write a simple which serves as instruction of what kind image edit you want to initiate.
4. Check the Sampling Set the settings as seen on the image.
5. Check the Output

Examples

IP Creation

Novel View Synthesis

Avatar Creator

Object Add

Object Removal

Object Replace

Background Swap

Virtual Try-On

Text Editing

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5510 members

Discord

Tired of dealing with complex setups and high memory demands just to make a single video? WAN2.2 Rapid-AllInOne changes the game by combining the best of WAN 2.2 and its accelerators into one lightweight, user-friendly model that works fast—even on lower VRAM systems.

Making AI videos usually means downloading multiple huge model files, dealing with memory errors, and babysitting complicated setups. WAN 2.2 Rapid-AllInOne rolls everything into a single file that runs faster and needs less VRAM.

It combines WAN 2.2, its accelerators, CLIP, and VAE into one model created by Phr00t. You load one file instead of several, it runs in 4 sampling steps, and works on lower-end hardware. Whether you're starting from an image or a text prompt, you get smooth motion without the usual technical headaches.

What we'll cover:

1. Why this model is easier than regular WAN 2.2
2. How to install and set up the workflow
3. Running different generation modes (image-to-video, text-to-video, etc.)
4. Examples and troubleshooting

Why Use Wan 2.2 Rapid-AllInOne?

Rapid-AllInOne is a fast, all-in-one AI video generation model developed by the creator “Phr00t,” who merged WAN 2.2 and various accelerators, along with CLIP and VAE, to deliver rapid, simplified video creation for image-to-video and text-to-video workflows. Its standout features are single-file convenience, suitable for low VRAM usage, and native integration of CLIP, VAE, and WAN accelerators for high performance and flexible output. The model is designed for speed, requiring only 4 sampling steps and 1 CFG, supporting dynamic tasks like last-frame and first-to-last-frame generation while ensuring compatibility with both WAN 2.1 and 2.2 LORAs.

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 1 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set Model Set the exact model as seen on the image. Use only the rapid mega model which can handle a different types of generation mode.
2. Load Input Load an input image. You can bypass any of these nodes depends of what type of generation mode you are using.
3. Write Prompt Write the required prompt. It is recommended to write a detailed prompt.
4. Check Sampling Check the sampling settings. Set it based from the recommended settings of the image.
5. Check Output

EXAMPLES

Image to Video

Text to Video

Last Frame Video

First - Last Frame Video

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5573 members

Discord

Generating Qwen images with Controlnet unlocks a powerful way to guide your AI creations using visual structure, lines, and forms drawn or extracted from reference images. Want better control over your AI image generation? Here's how to use Qwen Image with InstantX Union ControlNet to guide your creations with poses, edges, and depth maps.

With just a simple pose, edge, depth map, or quick sketch, you can shape exactly how your output looks. Whether you're working on precise designs or expressive portraits, this workflow gives you the control you need without the complexity.

Here's what we'll cover:
1. Why InstantX Union beats DiffSynth
2. Getting the workflow set up
3. Required models and custom nodes
4. Step-by-step walkthrough
5. Real examples and troubleshooting

Why InstantX Union is Better than DiffSynth?

InstantX Union ControlNet combines four control types (canny, soft edge, depth, and pose) into one model file. Instead of downloading separate models for each control type, you get everything in one package.

Unlike DiffSynth, which makes you load different models for different tasks, InstantX Union lets you switch between control types instantly. Less storage, less setup, same quality.

For Qwen Image users, this means creating complex, high-quality images with simpler setup, better compatibility, and instant access to the most common control modes all in a unified, user-friendly package.

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 4 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load Input Image Load an image. Image should be in good quality. Any resolution will do.
2. Set a Controlnet Set your desired controlnet based on your preferences. If there a human in the image you can use the pose.
3. Set the Models Set the exact models as seen on the image.
4. Write a Prompt Write a detailed of what of you kind of new image you want be in the input image.
5. Check Sampling Set the sampling as seen on the image.
6. Check Output

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5510 members

Discord

Ever found yourself wishing a portrait could actually speak, sharing stories with real movement and emotion? Now, that spark of imagination is within reach—no complicated setups required. With just a bit of creative input, you can watch your favorite images transform into lifelike, expressive talking portraits that surprise, engage, and even make you do a double-take.

What is InfiniteTalk? What are the Key Features?

InfiniteTalk is a powerful audio-driven video generation model designed to create unlimited-length talking avatar videos with exceptionally accurate lip sync, natural head and body movements, and stable facial expressions—all seamlessly aligning to input audio. What sets InfiniteTalk apart from MultiTalk is its enhanced stability, dramatically reduced distortions in hands and body, and superior lip synchronization, making each generated video look more realistic and less prone to awkward or exaggerated motion.

InfiniteTalk is an audio-driven video generation model that creates realistic talking avatar videos from static images or existing videos. It provides:

• Unlimited video length - Generate videos of any duration
• Accurate lip sync - Audio perfectly matches mouth movements
• Natural motion - Realistic head and body movements
• Multi-person support - Handle multiple speakers in one video
• Enhanced stability - Reduced distortions compared to MultiTalk

Perfect for content creators, educators, marketers, and developers who need professional talking avatars.

If you thought only big studios could achieve this kind of realism, prepare to be amazed—InfiniteTalk Video to Video hands you the power to let your portraits do the talking!

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 6 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set the Models Set the models as seen on the image. If you got an out of memory, enable the low mem load and use the fp8 version of models.
2. Set the Input Set the input as seen on the image. If you got a higher machine than Ultra then you can set it to the 720p resolution.
3. Load the Audio Load an audio file and should be a high quality audio.
4. Load the Video Load a video. If the video is vertical then set to vertical input dimesion. Otherwise, use the horizontal settings.
5. Write the Prompt You can write only a simple prompt.
6. Check the Sampling Set the sampling as seen on the image.
7. Check the Output

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5510 members

Discord

Uni3C is a ComfyUI model by Alibaba that converts static images into dynamic videos by transferring camera movements from reference videos. This tutorial covers complete setup and usage.

What is Uni3C?

Uni3C is a unified 3D-enhanced framework integrated into ComfyUI that enables precise, simultaneous control over both camera motion and human animation within video generation workflows. By leveraging a lightweight plug-and-play control module, Uni3C extracts and transfers motion—such as camera movements and character actions—from reference videos directly onto new scenes or images, eliminating the need for complex manual rigging or joint annotation.

This technology is especially valuable for digital artists, animators, filmmakers, virtual avatar creators, educators, and anyone in content creation seeking to bring static visuals to life with realistic, controllable movement—all with the creative freedom and modular workflow of ComfyUI.

Whether you’re just starting out or looking to elevate your animation game, let’s explore together how a few simple steps can transform your scenes from static to spectacular!

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 6 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Set the Height, Weight and Frames Set the frames up to 125. You can set the resolution up to 720 or 1080.
2. Load an Input Video for Control Reference Load an any type of video. The video must have a scene of camera movement.
3. Load an Input Image Upload an image which serves as the base image for the generation. It works with no subject or even in multiple subject.
4. Set the Models Set the models as seen on the image.
5. Write Prompt and Check Sampling Write a simple prompt which support the kind of camera or depth movement. Check the sampling settings as seen on the image.
6. Check the Output

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

What This Workflow Does

This ComfyUI workflow creates smooth animations by:

• Taking your starting image
• Generating an end frame with AI
• Creating seamless transitions between both frames
• Maintaining consistent subjects and backgrounds throughout

This ComfyUI workflow creates smooth animations by taking your starting image, generating an AI-powered end frame, and creating seamless transitions between both frames while maintaining consistent subjects and backgrounds throughout. The result is professional-quality animations with cinematic transitions, subject consistency across all frames, and context-aware scene evolution without jarring cuts or morphing artifacts.

Think of it as directing a movie scene where you define the beginning and ending poses, and the AI fills in all the natural movement between them. Whether you want a mother lifting her baby into an embrace, a flower blooming to reveal something magical, or a peaceful landscape transforming dramatically, this workflow handles the complex interpolation while keeping everything visually coherent.

Why is VACE and Flux Kontext has the advantage?

Integrating VACE - a powerful, unified model that brings precision, subject consistency, and cinematic quality to each frame, enabling coherent background changes and nuanced transformations. Including Flux Kontext, a context-aware image editing model, further refines this process by allowing detailed, prompt-driven modifications and maintaining object integrity or stylistic adjustments from the start to the end frame.

By combining VACE and Flux Kontext in the Wan Start-End Frame workflow, users gain the ability to craft sophisticated, professional-grade animations with smooth transitions, reliable subject identity, and creatively controlled scene evolution that simply isn’t possible with basic interpolation alone.

Let’s the journey toward truly seamless storytelling.

Download Workflow

Getting started requires downloading the workflow file and setting up the necessary components. First, download the workflow file and open ComfyUI, either locally or through ThinkDiffusion. Simply drag the workflow file into the ComfyUI window to load it.

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 9 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load First Frame Image Input your image as the first frame. You can choose either horizontal scale or vertical image.
2. Generate End Frame Image Set your end frame image by generating it using the flux kontext prompt.
3. Set Models Set the models as seen on the image. If you experience an out of memory reduced the weight more to fp8.
4. Write Prompt for First and End Frame Write a complete prompt here for your Start and End frame. It should be detailed and describes how does animation looks from the start to end.
5. Check Sampling Set the sampling settings as seen on the image.
6. Check Output

EXAMPLES

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

If you're having issues with workflow and visit us here at Discord #Help Desk or you may opt to email us at [email protected]

Join the ThinkDiffusion Discord Server!

ThinkDiffusion is your Stable Diffusion workspace in the cloud with unrestricted, bleeding edge opensource AI art tools. | 5510 members

Discord

Transform static portraits into realistic talking videos with perfect lip-sync using MultiTalk AI. No coding required.

Difficulty: Beginner-friendly
Setup Time: 15 minutes

What You'll Create

Turn any portrait - artwork, photos, or digital characters - into speaking, expressive videos that sync perfectly with audio input. MultiTalk handles lip movements, facial expressions, and body motion automatically.

Example Results:

• Portrait paintings that recite poetry
• Character artwork that delivers dialogue
• Profile photos that sing songs
• Multiple characters having conversations

In this tutorial, you’ll discover a surprisingly easy way to bridge the gap between static images and expressive animation. Whether you’re looking to enhance your social media posts, create memorable content for friends, or explore new storytelling techniques, this guide will open the door to a world where your characters can truly interact and entertain.

What is MultiTalk?

MultiTalk is an open-source AI framework that converts static images into realistic talking videos using audio input. Built by MeiGen AI, it accurately syncs lip movements and facial expressions to speech or singing, supporting both single and multi-person scenes.

With support for single or multi-person scenes, text prompts for emotion and behavior control, and compatibility with real or stylized characters, MultiTalk offers incredible creative flexibility. Integrated into ComfyUI and optimized for fast performance, it’s ideal for digital artists, content creators, educators, and developers who want to bring portraits, avatars, or original characters to life in seconds.

MultiTalk Framework

Let’s explore how you can turn imagination into motion, and watch as your creative visions become animated realities! The possibilities are as limitless as your imagination, so let’s get started and see where your characters can take you next!

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 8 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load an Image Load an input image. It should be clear and high in quality.
2. Set the Models Set the models as seen on the image. Don't change any setting because it may lead to out of memory. This workflow settings is already on the edge.
3. Write a Prompt Write a prompt. Include a word detailz, this is a trigger word for lora.
4. Check Sampling Check the sampling the settings. Set the steps count as it is because it uses a lora. Set the setting as seen on the image.
5. Check Output

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Flux Kontext expands your images into panoramic views directly in ComfyUI. Instead of cropping or stretching, it intelligently generates new content that extends beyond your image borders, creating seamless panoramic scenes.

What You'll Get

This workflow takes a standard image and generates extended panoramic versions by:

• Analyzing your input image's context and style
• Generating new content that naturally extends the scene
• Creating horizontal panoramas up to 3x the original width
• Maintaining consistent lighting, perspective, and artistic style

Best for: Landscapes, cityscapes, interior shots, and any scene where you want to reveal "what's beyond the frame."

What is Image-to-Panorama?

A panorama is an image that captures a much wider field of view than a standard photograph, often stretching across an entire landscape or environment to reveal far more than what a single frame can show. Traditionally created by stitching together multiple overlapping photos, panoramas provide a seamless, immersive scene that can be horizontal, vertical, or even a full 360-degree view. In image generation, panoramas are especially useful because they allow creators to expand the context and storytelling potential of a single image, offering a broader perspective and more detail. This capability is valuable for photographers and artists seeking to produce striking, high-resolution visuals, as well as for architects and real estate professionals who need to showcase spaces in a comprehensive, interactive way.

Marketers, educators, and those in travel or tourism can also benefit, using image-to-panorama and 3D view features to create virtual tours, interactive content, and engaging educational materials. Ultimately, panoramas and 3D views transform ordinary images into immersive experiences, making them more informative, captivating, and useful for a wide range of creative and professional applications.

Get ready to see your pictures in a whole new light!

Download Workflow

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 8 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load an Image Load an image that is good for 3D panorama view somthing like surrounding, landscape, interior, etc.
2. Set the Models Set the models as seen on the image.
3. Check Prompt Check the prompt. You don't need to write anymore.
4. Check Sampling Check the sampling settings it should be the same as seen on the image.
5. Check Crop and Panorama Settings Don't change the settings for crop and other settings.
6. Check Upscale Check the upscale settings. It should x2 only for upscale. Otherwise the workflow will crash.
7. Check Output

Examples

Troubleshooting

Red Nodes: Install missing custom nodes through ComfyUI Manager
Out of Memory: Use smaller expansion factors or switch to Ultra machine
Poor Quality: Check input image resolution and adjust kontext strength
Visible Seams: Lower strength and ensure good prompt description

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

Want to create custom AI image models but find the process intimidating? This guide shows you how to train your own LoRA models using FluxGym - no coding experience required.

Whether you want to generate images in a specific art style, create consistent characters, or adapt AI models for your unique needs, you'll learn everything you need to know in about an hour.

Let’s dive in and unleash your creative potential with FluxGym!

What is LoRA?

LoRA, or Low-Rank Adaptation, is a technique in artificial intelligence that enables efficient and targeted fine-tuning of large pre-trained models, such as those used in image generation or language processing-without the need to retrain or modify the entire model.

Think of it as teaching an existing AI model new tricks—like recognizing a specific art style or character—while keeping all its original knowledge intact.

This makes LoRA perfect for:

• Creating images in a particular artistic style
• Generating consistent characters for stories or projects
• Adapting AI models for specific visual themes
• Customizing outputs without massive computing resources

What is FluxGym?

FluxGym is an open-source, web-based tool designed to make training Flux LoRA models accessible and straightforward, especially for users with limited hardware resources. Created by the developer behind Pinokio, it stands out for its user-friendly Gradio interface.

Key features:

• Easy-to-use web interface
• Real-time training progress tracking
• Automatic sample image generation
• Works with limited hardware resources
• No command line or coding required

How to Train a Flux LoRa using FluxGym?

The first part is preparing your dataset.

Tips for Datasets

• Use short, descriptive filenames for your images.
• Use high-quality, sharp images (ideally 1024x1024) with the subject clearly visible and centered.
• Maintain consistency in style, lighting, and subject focus across your dataset.
• Include variety by showing the subject in different poses, angles, and backgrounds to help the model generalize.
• Crop to a square aspect ratio and organize images in a clearly named folder.
• Write accurate captions for each image, describing the subject and style; review auto-generated captions for accuracy.
• Aim for quality over quantity-20–30 well-chosen images are often better than a larger, inconsistent set.

What you'll need?

- 20-30 high-quality images (1024x1024 recommended)
- Access to FluxGym (via ThinkDiffusion or local install)
- About 2 hours+ for training (depends with no. of dataset)
- Basic understanding of your creative goal

Procedures

Step 1

Step 2

Step 3

Examples of LoRA

Rodrigo Duterte (Character LoRA)

Dan Mumford Art (Style LoRA)

Sarah Duterte (Character LoRA)

Anato Finnstark Art (Style LoRA)

Once you're comfortable with character and style LoRAs, you can experiment with:

• Object LoRAs: For specific items or props
• Pose LoRAs: For particular poses or compositions
• Modifier LoRAs: For specific effects or modifications

Each type serves different creative needs and can be combined for even more precise control over your AI-generated images.


If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

If you're ready to try this yourself but need more powerful hardware, ThinkDiffusion offers FluxGym access through your browser with high-end GPUs.

The key to success is starting simple: pick one clear goal (like a specific art style), gather quality images, and follow the steps carefully. With a bit of practice, you'll be creating custom AI models that perfectly match your creative vision.

If you enjoy ComfyUI and you want to test out creating awesome animations, then feel free to check out this Flux LoRA Training using ComfyUI. And have fun out there with your noodles!

Flux Kontext lets you edit images precisely using text descriptions. Tell it what to change, and it modifies only those specific parts while keeping the rest intact. This is a gamechanger for control, and delivers incredible quality.

What is Flux Kontext?

Flux Kontext is an AI image editing model by Black Forest Labs that excels at targeted modifications. Instead of generating entirely new images, it edits existing ones based on your text instructions.

Core capabilities:

• Local editing: Change specific parts without affecting the whole image
• Character consistency: Keep people looking the same across multiple edits
• Style transfer: Apply artistic styles to existing images
• Multi-round editing: Make several edits in sequence
• Object manipulation: Add, remove, or modify objects

Available versions: Pro, Max, and Dev. This guide covers Flux1-kontext-dev, which is freely available under a non-commercial license

Multi-Round Editing (Iterative)

How Multi-Round Editing Works

Flux Kontext's strength is iterative editing. You can:

1. Start with one image
2. Make a targeted edit
3. Use the result for the next edit
4. Continue refining step by step

Each edit builds on the previous one while maintaining visual consistency and quality.

Limitations of Flux Kontext

• Quality degradation: After 6+ iterative edits, images may show artifacts
• Instruction following: Sometimes misses specific prompt details
• Limited world knowledge: May struggle with contextually accurate content
• Distillation artifacts: Processing can introduce visual issues
  
  Get ready to unlock a smarter, sharper, and more creative approach to image generation—because your next masterpiece is just around the corner!

How to Use Flux Kontext in ComfyUI

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:
   • ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 4 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes
1. Load an Image Load an image. Make sure that it is clear and free from artifacts and blurred. If you want to combine the two image, just enable the 2nd load image.
2. Set the Models Set the models as seens on the image.
3. Write a Prompt Write a simple prompt which describes the noun and you want to be for the output. See the examples below for your guidance.
4. Check the Sampling Set the settings as seen on the image. Do not change the CFG higher as it may cause artifacts.
5. Check the Output

Examples

Character Consistency

Add / Edit Text

Remove Objects

Style Reference

Switch View

Multiple Input Images

Change Light

Image Editing

Restyle

Tips for Better Results

1. Start simple: Begin with basic edits before complex changes
2. Be specific: Clear descriptions work better than vague ones
3. Check quality: Monitor for artifacts after each edit
4. Limit iterations: Avoid more than 5-6 sequential edits
5. Use good source images: High-quality inputs produce better outputs

Troubleshooting

• Red nodes: Install missing custom nodes via ComfyUI Manager
• Model errors: Verify all 4 models are downloaded correctly
• Poor results: Simplify prompts and retry
• Artifacts: Reduce CFG settings or start with fresh image

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.

MAGREF lets you create videos from multiple reference images while keeping each person or object looking consistent throughout the video. Instead of generating random characters, you can use your own photos to control exactly who or what appears in the final video. This guide shows you how to set up and use MAGREF in ComfyUI to create videos with multiple subjects that maintain their original appearance.

What is MAGREF?

MAGREF, or Masked Guidance for Any-Reference Video Generation, is a diffusion-based AI framework that generates videos from multiple reference images while preserving subject identity. It uses region-aware dynamic masking to handle any number of subjects and pixel-wise channel concatenation to maintain fine details. This makes it especially useful for creating videos where specific people, objects, or backgrounds need to appear exactly as they do in your source images.

MAGREF is especially valuable for creators, animators, and researchers seeking to produce customizable, multi-subject videos with exceptional fidelity and control, making it a standout solution for both creative and professional video generation needs.

Types of MAGREF

MAGREF offers three flexible video generation modes: 

• Single ID, which uses one reference image to keep a single subject consistent throughout the video; 
• Multi-ID, which allows multiple subjects from different reference images to appear together while maintaining their unique identities;
• ID-Object-Background, which lets users combine references for people, objects, and backgrounds to create complex, multi-layered scenes. These options make MAGREF suitable for everything from simple, personalized videos to rich, detailed multi-subject compositions.

With just a few easy steps, you’ll see how effortless and fun it can be to bring your memories to life in ways you never thought possible!

How to Use Wan MAGREF for Video Generation

Installation guide

1. Download the workflow file
2. Open ComfyUI (local or ThinkDiffusion)
3. Drag the workflow file into the ComfyUI window
4. If you see red nodes, install missing components:

• ComfyUI Manager > Install Missing Custom Nodes

Custom Nodes

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

1. Go to the ComfyUI Manager  > Click Install Missing Custom Nodes

2. Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Required Models

For this guide you'll need to download these 5 recommended models.

1. Go to ComfyUI Manager  > Click Model Manager

2. Search for the models above and when you find the exact model that you're looking for, click install, and make sure to press refresh when you are finished.

If Model Manager doesn't have them: Use direct download links (included with workflow) and upload through ThinkDiffusion MyFiles > Upload URL. Refer our docs for more guidance on this.

You could also use the model path source instead: by pasting the model's link address into ThinkDiffusion MyFiles using upload URL.

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

There are 2 Group Nodes that you need to remember.
You can disable first the 2nd Group Node, so you can test first the brightness or saturation each of the input image of 1st Group node before you can run full run of workflow. Different tone of brightness/saturation of each image may lead to uneven color at the output.

Steps	Recommended Nodes
1. Load Image (1st Group) Load a high quality and should not be pixelated and blurred. Just follow the settings as seen on the image.
Check Bridge Settings (optional) This area is optional on which you set on what type of MAGREF you want to use. If you'll a 1 image only, then use a 1 route bridge only and connect to Resize Image and Get Image Size node. If you need multiple, then use the Image Concatenatte Multi node and reconnect the connection according to your number to inputs and reconnects to the other side.
2. Set Input Settings (2nd Group) Set the input settings for the image. The model is only compatible to 480p and 720p. Take note that frames should be higher than 81 as it may crashed if goes beyond it.
3. Set Models (2nd Group) Set the models as seen on the image.
4. Write Prompt (2nd Group) Write a simple prompt and you don't need to be specific and super detailed. Write that describes the noun and add some simple action words.
5. Check Sampling Settings (2nd Group) Check the samplings settings as seen on the image. Steps should be only 4-6 and CFG should 1-2 only and nothing else.
6. Check Output (2nd Group) Check the generated output. Output may not perfect at one test. If you're not okay with result, just tweak the prompt and rerun the workflow.

Examples

Single ID Video

Multi-ID Video

ID + Object + Background

If you’re having issues with installation or slow hardware, you can try any of these workflows on a more powerful GPU in your browser with ThinkDiffusion.