The ability to transfer poses seamlessly between different characters is a valuable asset. Traditionally, achieving this involved complex 3D software and time-consuming processes. However, with the advent of OpenPose and its integration with Stable Diffusion, a revolutionary solution has emerged. In this crash course, we delve into the remarkable capabilities of OpenPose and how it synergizes with Stable Diffusion, opening up new possibilities for character generation.

1. Introduction OpenPose Crash Course

To facilitate the smooth transfer of character poses using ControlNet and the OpenPose ControlNet model, our initial focus will be on the installation of the ControlNet Extension. This marks the first step in our two-phase installation process.

Detailed step-by-step instructions for the installation of the ControlNet Extension are provided below. The following chapter will delve into the installation process of the OpenPose ControlNet Model.

### ControlNet Extension Installation
- Launch Stable Diffusion
- Click on "**Extensions**" in the menu
- Select "**Install from URL**"
- Paste the Official ControlNet Github URL: https://github.com/Mikubill/sd-webui-controlnet
- Click on "**Install**" to add the ControlNet Extension

![stable-diffusion-openpose-course_download_controlnet_extension_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702309167/stable_diffusion_openpose_course_download_controlnet_extension_nextdiffusion_59eac089ee.webp)

After installing the ControlNet Extension, navigate to the "**Installed**" tab, and select "**Apply and restart UI**". Additionally, I highly recommend performing a complete restart of your Stable Diffusion for a seamless experience and to address any potential errors.

### ControlNet Extension Stable Diffusion Web UI
Once the Stable Diffusion web UI has restarted, you will notice a new expansion panel when scrolling down in the interface. Here, you should find the recently added ControlNet Extension. It should look something like this when the ControlNet expansion panel is opened:

![stable-diffusion-openpose-course_download_controlnet_extension_stablediffusion_interface_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702335814/stable_diffusion_openpose_course_download_controlnet_extension_stablediffusion_interface_nextdiffusion_0890d97512.webp)

### ControlNet Extension Folder Location
Upon completing the installation of the ControlNet Extension, a new folder should become visible within the following file path: "**stable-diffusion-webui\extensions**". It is at this location that the recently generated "**sd-webui-controlnet**" folder will be present.

2. Installation ControlNet Extension

Now that the ControlNet Extension is set up, the next step involves configuring the ControlNet Models, with a specific focus on installing the OpenPose ControlNet Model. Follow the steps below for this process:

- Go to [Official ControlNet Models](https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main) Page
- Download the OpenPose ControlNet Model named: "**control_v11p_sd15_openpose.pth**"

### OpenPose ControlNet Model File Location
Once downloaded, place the model in following folder location: "**stable-diffusion-webui-extensions/sd-web-ui/ControlNet/models**" within the Stable Diffusion folder.

![stable-diffusion-openpose-course-controlnet_models-file-location.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702412913/stable_diffusion_openpose_course_controlnet_models_file_location_873f176fa0.webp)

### Pre-trained ControlNet Models (.safetensors files)
You also have the option to download the [.safetensors pre-trained model](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/tree/main). Again place the "**control_v11p_sd15_openpose_fp16.safetensors**" file within the correct folder as described above.

_I would personally suggest downloading the .safetensors version, as it occupies less storage space and can't contain malicious code._

### Pairing ControlNet Models with Preprocessors
Each model needs to be paired with the appropriate preprocessor. In this course, we will utilize the OpenPose pre-processor. Ensure that you associate it with either the original or pre-trained OpenPose model within the Stable Diffusion web UI. If the recently downloaded model does not appear immediately, you can click the blue refresh icon to potentially display the newly downloaded ControlNet model.

![stable-diffusion-openpose-course-pre-processor-paired-models-nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702414195/stable_diffusion_openpose_course_pre_processor_paired_models_nextdiffusion_71e3bd1757.webp)

_Typically, selecting a specific "Control Type" will automatically load the appropriate model. However, it's advisable to keep this in mind._

3. Installation OpenPose ControlNet Model

In the upcoming section, we will delve into the dynamic interaction between ControlNet and OpenPose. Having successfully installed and configured both the ControlNet Extension and the OpenPose Model, we are poised to embark on a journey through practical examples. Our focus will be on uncovering the optimal configurations that enable the creation of captivating characters exhibiting a diverse range of poses.

### Setting the Stage for Character Poses
To get started, it's important to decide what poses you want for your characters. In this chapter, it's a good idea to find pictures of characters in poses you like. These images will be really helpful as we go through the examples and show you how to easily transfer the poses you have in mind. We have downloaded the following full body poses from the internet:

![stable-diffusion-openpose-course-exploring-controlnet-openpose-nextdiffusion-body-poses.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702484891/stable_diffusion_openpose_course_exploring_controlnet_openpose_nextdiffusion_body_poses_359517e6ce.webp)

### ControlNet Settings
Let's start exploring ControlNet & OpenPose by uploading our first image on to the ControlNet unit and do the following:

- Enable ControlNet by ticking the "**Enable**" checkbox
- Check the "**Pixel Perfect**" checkbox
- Select "**OpenPose**" as the Control Type
- Select "**openpose_full**" as the Preprocessor (since it is a full body portrait)
- Select "**Control_v11p_sd15_openpose**" as the Model

![stable-diffusion-openpose-course-exploring-controlnet-openpose-nextdiffusion-example-1.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702485278/stable_diffusion_openpose_course_exploring_controlnet_openpose_nextdiffusion_example_1_ac65c6cda8.webp)

Navigate back to the top and select a checkpoint. In this case, we'll choose the "**Realistic Vision**" checkpoint for a lifelike result. Formulate your prompts and make sure the width and height align with the original image you dropped onto the ControlNet canvas. It's now time to click on the generate button to produce images using the pose from your initial picture.

### Example 1

![stable-diffusion-openpose-course-exploring-controlnet-openpose-nextdiffusion-example-1-final.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702487077/stable_diffusion_openpose_course_exploring_controlnet_openpose_nextdiffusion_example_1_final_1631024cd0.webp)

### Example 2

![stable-diffusion-openpose-course-exploring-controlnet-openpose-nextdiffusion-example-2-final.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702487524/stable_diffusion_openpose_course_exploring_controlnet_openpose_nextdiffusion_example_2_final_1b08f9cfff.webp)

### Example 3

![stable-diffusion-openpose-course-exploring-controlnet-openpose-nextdiffusion-example-3-final.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702487742/stable_diffusion_openpose_course_exploring_controlnet_openpose_nextdiffusion_example_3_final_d9bad29029.webp)

### Prompts & Settings
- **Positive Prompt**: 30 year old woman, white silk dress, red lips, cleavage, outdoor, tropical plants in background, natural light, perfect light, perfect shadows, facing camera, looking in camera
- **Negative Prompt**: (worst quality:1.2), (low quality:1.2), (lowres:1.1), (monochrome:1.1), (greyscale), multiple views, comic, sketch, (((bad anatomy))), (((deformed))), (((disfigured))), mutation hands, mutation fingers, extra fingers, missing fingers, watermark
- **Sampling Steps**: 30
- **Sampler**: DDIM
- **CFG scale**: 7
- **Size**: 560x904
- **Model**: [Realistic Vision](https://civitai.com/models/4201/realistic-vision-v60-b1)

### Chapter Conclusion
In this chapter, we've emphasized that ControlNet, combined with the OpenPose Model, serves as a straightforward and efficient tool for translating poses from an initial image to your desired output. This involves providing a clear prompt and choosing the appropriate checkpoint. Moving forward, the next chapter will explore a resource where we can discover high-quality poses, along with delving into additional settings offered by ControlNet.

4. Exploring ControlNet & OpenPose

In the previous chapter, we placed an initial image featuring a full body pose onto the control canvas and opted for the "**openpose_full**" preprocessor, given that it was a portrait capturing the entire body.

![stable-diffusion-openpose-course-pre-processor-paired-models-nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702414195/stable_diffusion_openpose_course_pre_processor_paired_models_nextdiffusion_71e3bd1757.webp)

However, there is a diverse range of preprocessors to choose from:

- none
- dw_openpose_full
- openpose
- openpose_face
- openpose_face_only
- openpose_full
- openpose_hand

The names of the preprocessors usually suggest their image processing methods. However, let's illustrate this with some examples using the following image, which will undergo processing with various preprocessors:

![girl-brunette-gorgerous.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702570195/girl_brunette_gorgerous_f737dba14b.webp)

### Preprocessor "dw_openpose_full"

For analyzing a full-body portrait, "**dw_openpose_full**" is the top choice. It processes the face, body, and hands effectively.

![stable-diffusion-openpose-course_preprocessor_dw_openpose_full (5).webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702572291/stable_diffusion_openpose_course_preprocessor_dw_openpose_full_5_d4c64b91d4.webp)

### Preprocessor "openpose"
The "**openpose**" preprocessor will simply render the image as a basic stick figure, focusing solely on the body and not processing the hands or face. This suggests that the hands can be generated in any direction, but the results may not be as optimal as when the hands are processed.

![stable-diffusion-openpose-course_preprocessor_openpose.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702570913/stable_diffusion_openpose_course_preprocessor_openpose_8e0d1788ce.webp)

### Preprocessor "openpose_face"
The "**openpose_face**" preprocessor will process the image, concentrating on both the body and face while neglecting the hands. Once more, this indicates that the hands may be generated in any direction, but the outcomes might not be as effective as when the hands undergo preprocessing.

![stable-diffusion-openpose-course_preprocessor_openpose_face.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702571140/stable_diffusion_openpose_course_preprocessor_openpose_face_73ed60b18e.webp)

### Preprocessor "openpose_faceonly"
The "**openpose_faceonly**" preprocessor exclusively focuses on the face, excluding the body and hands. It's the ideal option when you specifically desire the facial outline and nothing more. However, keep in mind that the generation of body and hand poses may vary significantly.

![stable-diffusion-openpose-course_preprocessor_openpose_faceonly.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702571458/stable_diffusion_openpose_course_preprocessor_openpose_faceonly_33c2e0a010.webp)

### Preprocessor "openpose_full"
The "**openpose_full**" preprocessor handles everything—hands, face, and body—similar to the "**dw_openpose_full**" preprocessor. However, for finer details, I recommend opting for "**dw_openpose_full**" as it typically processes images with more precision and accuracy.

![stable-diffusion-openpose-course_preprocessor_openpose_full.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702571777/stable_diffusion_openpose_course_preprocessor_openpose_full_9e65166671.webp)

### Preprocessor "openpose_hand"
The "**openpose_hand**" preprocessor accentuates hands and body while excluding the face. In most instances, you may notice subtle variations in facial features compared to the original, although it attempts to replicate the face's angle.

![stable-diffusion-openpose-course_preprocessor_openpose_hand.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702572050/stable_diffusion_openpose_course_preprocessor_openpose_hand_c347e14e45.webp)

### The "none" Preprocessor
There is one last preprocessor to explain, which is the "**none**" preprocessor. As the name implies, it involves minimal processing. When choosing the "none" preprocessor ensure you drop an image onto the canvas that's already processed, like stick figure images that have undergone prior processing. Below is an example of how the settings should look when opting for the "none" preprocessor.

![openpose_crash_course_openpose_preprocessors_none_option_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702633504/openpose_crash_course_openpose_preprocessors_none_option_nextdiffusion_6e55634b41.webp)

### Conclusion Preprocessors
In summary, we've explored various preprocessors, each tailored to specific aspects of portrait analysis. It's essential to understand that every preprocessor has different use cases. Before selecting one, clarify your preferences to achieve the desired outcome. With this knowledge and practical examples, you're well-prepared to effectively use these preprocessors.

5. OpenPose Preprocessors

In this section, we will delve into modifying the pre-processed stick figure. This can prove beneficial if you find yourself dissatisfied with how the preprocessing has interpreted your image, or if you wish to adjust the pose according to your preferences. This allows you to make subtle adjustments to the pose before generating image outputs. Let's start this process of with dropping a newly downloaded image onto the canvas, selecting the "**OpenPose**" Control Type followed by selecting an appropriate preprocessor and then clicking the "run preprocessor" icon next to the selected preprocessor. This will show a newly processed stick figure on the right canvas.

![openpose_crash_course_openpose_editing_stick_figure_example_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702755618/openpose_crash_course_openpose_editing_stick_figure_example_nextdiffusion_49f6fbb6ea.webp)

In the bottom right corner of the stick figure you have 3 options to choose from:

- Edit (Which we will explain in this section)
- JSON (Which will download the stick figure as a JSON file)
- Close (Which will remove and hide the preview)

### Editing Stick Figure
Let's initiate the editing process by selecting the "**edit**" button. Once clicked, a new window will emerge. Within this window, the editing options are situated on the left side, while the right side displays preview images, including the stick figure.

![openpose_crash_course_openpose_editing_stick_figure_window_example_nextdiffusion (1).webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702755764/openpose_crash_course_openpose_editing_stick_figure_window_example_nextdiffusion_1_90a08c52d3.webp)

I suggest utilizing the editing tool primarily for making minor adjustments or adding missing elements to the stick figure, especially in cases where the image hasn't been properly preprocessed. In our scenario, all elements are accurately processed, but we aim to make a slight adjustment to her pose.

To achieve this, simply scroll down within the left editing section until you reach the "**Pose Control**" section. Within this segment, you'll find the expansion panel for the preprocessed person, labeled "Person 1" in our example. If multiple individuals were processed, you will observe multiple dropdowns, each corresponding to a distinct person.

![openpose_crash_course_openpose_editing_stick_figure_window_pose_control_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702747572/openpose_crash_course_openpose_editing_stick_figure_window_pose_control_nextdiffusion_6a8b0abfcf.webp)

Within the "**Person 1**" dropdown menu, you can unveil keypoints that weren't initially processed or choose to conceal specific parts. Just click the "eye" icon next to the body part to switch visibility. In our instance, the majority of body parts were effectively processed, as seen in the precise representation of both hands and the face.

However, for this example, I'd like her to have both hands on her hips. To achieve this, let's begin by grouping her right hand for flexible movement. Then, we'll ensure her right elbow and right wrist are perfectly aligned by making adjustments on the canvas. Our final adjustments should look something like:

![openpose_crash_course_openpose_editing_stick_figure_window_pose_control_nextdiffusion_.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702757209/openpose_crash_course_openpose_editing_stick_figure_window_pose_control_nextdiffusion_2dda4779b1.webp)

Once all adjustments are complete, scroll back to the top and select "**Send pose to ControlNet**". Ensure you use the same dimensions as the edited canvas, pick a checkpoint, write a prompt, and click generate to witness stunning image generations reflecting the pose you just crafted.

_For additional details on certain settings, I recommend watching the video._

### Examples Hands on Hips Pose
Here are several examples generated using a cartoon-style checkpoint.

![openpose_crash_course_openpose_editing_stick_figure_final_examples_hands_on_hip_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702770168/openpose_crash_course_openpose_editing_stick_figure_final_examples_hands_on_hip_nextdiffusion_11c613c3b1.webp)

- **Positive Prompt**: 30 year old woman, red dress, red hair, cleavage, outdoor, tropical plants in background, natural light, perfect light, perfect shadows, facing camera, looking in camera
- **Negative Prompt**: (worst quality:1.2), (low quality:1.2), (lowres:1.1), (monochrome:1.1), (greyscale), multiple views, comic, sketch, (((bad anatomy))), (((deformed))), (((disfigured))), mutation hands, mutation fingers, extra fingers, missing fingers, watermark

### Conclusion Editing Stick Figures
In conclusion, our exploration delved into the practical aspect of editing preprocessed stick figures, offering users the valuable capability to fine-tune poses according to their unique preferences. This editing feature serves as a crucial tool in the creative process, allowing users to refine outcomes and achieve desired poses before generating the final image. The seamless integration of preprocessing and editing significantly enhances user control, ensuring a more personalized and satisfying experience in the creative journey.


6. Editing Stick Figures

Now, let's explore a practical scenario where we will employ OpenPose along with the image-to-image inpainting feature to alter the clothing style. In this case, we'll utilize OpenPose to maintain the pose integrity and ensure that our clothing aligns seamlessly with the established pose.

### Img2Img Settings
Let's initiate this practical use case by undertaking the following steps:

- Open the Stable Diffusion interface.
- Click on the "**img2img**" tab.
- Identify and choose the "**Inpaint**" sub-tab among the available options.

Let's proceed by placing the initial image onto the canvas and create a mask to delineate the specific areas we intend to modify, providing a targeted and precise approach to the alterations we are about to make. This step involves defining regions within the image that will undergo changes, ensuring a nuanced and controlled transformation.

![openpose_crash_course_openpose_inpainting_inpainted_example_ui_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702816720/openpose_crash_course_openpose_inpainting_inpainted_example_ui_nextdiffusion_7226bc5a40.webp)

### Inpaint Settings (img2img)
To attain the desired result, simply follow these steps within the inpaint settings, which essentially consist of the default configuration:

- Mask blur : 4
- Mask Mode: Inpaint Masked
- Masked Content: original
- Inpaint Area: Whole Picture

![openpose_crash_course_openpose_inpainting_inpainting_settings_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702816836/openpose_crash_course_openpose_inpainting_inpainting_settings_nextdiffusion_5445cb8035.webp)

Now let's move on to some more settings, which needs some altering:

- Sampling method: Euler a (This choice helps maintain image clarity)
- Sampling Steps: 30
- Match the width and height ratio with the original image for accurate proportions. Click the "autodetect" feature to transfer the dimensions.
- CFG scale: 7
- Denoising strength: 0.75
- Seed: -1

![openpose_crash_course_openpose_inpainting_inpainting_settings_2_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702817236/openpose_crash_course_openpose_inpainting_inpainting_settings_2_nextdiffusion_ca14dd45f0.webp)

### ControlNet Settings (OpenPose)
Scroll down the page to access the ControlNet panel. Begin by placing your initial image onto the canvas. Before you can place an image onto the canvas, make sure to check the "**Upload independent control image**" box. Ensure that you choose the "**OpenPose**" Control Type and opt for an appropriate preprocessor. In this instance, we will choose the "dw_openpose_full" preprocessor, given its ability to capture a full-body pose. Feel free to execute the preprocessor by clicking the icon and observe how your pose transforms when generating images.

![openpose_crash_course_openpose_inpainting_inpainting_settings_controlnet_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702817537/openpose_crash_course_openpose_inpainting_inpainting_settings_controlnet_nextdiffusion_e594ea52da.webp)

With all the required configurations accurately set, let's advance to the next stage. We'll choose the "[Realistic Vision Inpainting Model](https://civitai.com/models/4201?modelVersionId=245627)" and craft a positive prompt to change the character's outfit. Furthermore, we'll devise a contrasting negative prompt. Once all the remaining settings are properly adjusted, a simple click on "Generate" will allow us to witness the creation of remarkable outfit transformations.

### Example Outfits
![openpose_crash_course_openpose_inpainting_inpainting_examples_output_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1702821789/openpose_crash_course_openpose_inpainting_inpainting_examples_output_nextdiffusion_d316532a12.webp)

### Conclusion OpenPose & Inpainting Usecase
As we conclude our journey, we've effectively utilized the powerful features of Stable Diffusion, notably through the inpainting tab, in combination with ControlNet and OpenPose. This collaborative synergy ensures the accurate preservation of character poses, simplifying the process of incorporating clothing changes and adding a layer of simplicity to the overall transformation.

7. OpenPose & Inpainting

1. You must have Stable Diffusion Installed

By the end of the course you will master how to effortlessly transfer character poses using ControlNet and the OpenPose Model.

large_the-ultimate-stable-diffusion-openpose-course-nextdiffusion-next-diffusion-course.webp

medium_the-ultimate-stable-diffusion-openpose-course-nextdiffusion-next-diffusion-course.webp

small_the-ultimate-stable-diffusion-openpose-course-nextdiffusion-next-diffusion-course.webp

thumbnail_the-ultimate-stable-diffusion-openpose-course-nextdiffusion-next-diffusion-course.webp

the-ultimate-stable-diffusion-openpose-course-nextdiffusion-next-diffusion-course.webp

OpenPose Crash Course (Stable Diffusion)

Introducing the ReActor Face Swapping Extension—a powerful tool crafted to enhance your face-swapping experience. In this crash course, we'll swiftly guide you through the steps to download and leverage the ReActor extension within Stable Diffusion for achieving realistic face swaps. Explore additional details on the official [ReActor GitHub page](https://github.com/Gourieff/sd-webui-reactor) for in-depth information.

### Advantages of the ReActor Extension

- High-Resolution Face Swaps with Advanced Upscaling
- Optimized CPU Performance
- Compatibility Across SDXL and 1.5 Models
- Automatic Gender and Age Detection
- Uncensored Swaps (No NSFW Filter)
- Continuous Development and Updates

The ReActor Extension distinguishes itself by enabling high-resolution face swaps with advanced upscaling, optimized for CPU-only setups, providing compatibility across SDXL and 1.5 models. It streamlines face-swapping with automatic gender and age detection, supports uncensored swaps, and excels in continuous development, bringing forth evolving features and advancements in face-swapping technology.

### Ready to Begin?
Embark on our ReActor Face Swapping Course and experience the simplicity of face swapping within Stable Diffusion. In the upcoming chapter, we'll explore the installation process for the ReActor Extension. Let's start learning.

1. Introduction ReActor Extension

### Installing ReActor Extension

- On Windows, start by downloading and installing [Community Visual Studio](https://visualstudio.microsoft.com/downloads/).
- Open Stable Diffusion and navigate to the "**Extensions**" tab in the web UI.
- In the "**Install from URL**" tab, paste the following URL: https://github.com/Gourieff/sd-webui-reactor

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_extension_install.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704534032/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_extension_install_2b3498da8e.webp)

After the installation process is complete, go to the "**Installed**" section. Click on "**Apply and restart UI**" to initiate a User Interface restart, ensuring the successful integration of the extension.

### ReActor Extension Folder Location
Following a successful installation, locate the downloaded extension in the "**\stable-diffusion-webui\extensions**" folder. Look for a folder named "**sd-webui-reactor**" within this directory.

### Encountering ReActor Extension Installation Errors?
If you face any errors during the installation of the ReActor Extension, we recommend referring to the official [ReActor GitHub Page](https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#troubleshooting). It includes a comprehensive "**Troubleshooting**" section that provides explanations for common errors along with their solutions.

2. Installing the ReActor Extension

Upon successful installation, observe the appearance of the ReActor expansion panel in both the "**txt2img**", "**img2img**" and "**extras**" tabs within the Stable Diffusion UI.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704223549/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_717a6c13fc.webp)

The face-swapping process with ReActor involves a two-step approach. Initially, the AI image undergoes full rendering, after which the face from the ReActor image is applied. This distinction becomes apparent during rendering, where the image appears finalized but requires extra time for face integration. Let's explore this with an example!

- Navigate to the "**txt2img**" tab within Stable Diffusion.
- Launch the ReActor Expansion panel, and place a facial portrait onto the left canvas ("**Single Source Image**"). Ensure that it features only a single face. As an illustration, let's use the charming Angelina Jolie.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_enable_extension.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704224354/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_enable_extension_0369e49c02.webp)

- "**Enable**" ReActor by checking the checkbox
- Create both a Positive and Negative prompt
- Click "**Generate**"

**Note**: The initial face swap using the ReActor extension might have a longer duration as it unpacks essential models from a zip file. However, subsequent swaps will be quicker.

### Face Swapped Example (Angelina Jolie)
The final result is the generation of a striking face that resembles Angelina Jolie, crafted based on your provided input prompt.

![stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_example_angelina_jolie.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704279066/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_example_angelina_jolie_88e4e1f920.webp)

### Prompt & Settings
- **Positive Prompt**: 30 year old woman, white silk dress, red lips, waist and up, upper body, cleavage, outdoor, tropical plants in background, white villa background, natural light, perfect light, perfect shadows, facing camera, looking in camera, dynamic pose
- **Negative Prompt**: (worst quality:1.2), (low quality:1.2), (lowres:1.1), (monochrome:1.1), (greyscale), multiple views, comic, sketch, (((bad anatomy))), (((deformed))), (((disfigured))), mutation hands, mutation fingers, extra fingers, missing fingers, watermark
- **Sampling Steps**: 30
- **Sampler**: DDIM
- **CFG scale**: 7
- **Size**: 704x392
- **Model**: [Realistic Vision](https://civitai.com/models/4201/realistic-vision-v51)

3. Exploring the ReActor Extension

Enhance your face-swapped image using the ReActor Extension's upscale feature. To get started, go to the "**Upscale**" subtab within the ReActor Extension.

### ReActor Upscaling Settings
Choose to first swap faces and then upscale by selecting the checkbox (recommended for optimal results). Decide on your preferred upscaling method; I personally recommend "**R-ESRGAN 4x+**" for scaling up realistic images and "**R-ESRGAN 4x+ Anime6B**" for upscaling cartoon-styled images.

Here's an example of the settings, illustrating the appearance when scaling the face-swapped image with the R-ESRGAN 4x+ upscaler and scaling it by a factor of 2.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_upscale_feature.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704280970/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_upscale_feature_e63dc83094.webp)

### Upscaling Example (Angelina Jolie)
![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_upscale_example.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704281760/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_upscale_example_9aed8086f2.webp)

### Upscaled Images Output Location:

- When using the "txt2img" tab: The upscaled image will be visible in the default folder: "\stable-diffusion-webui\outputs\txt2img-images" upon image generation.
- When using the "img2img" tab: The upscaled image will be visible in the default folder: "\stable-diffusion-webui\outputs\img2img-images" upon image generation.
- When using the "extras" tab: The upscaled image will be visible in the default folder: "\stable-diffusion-webui\outputs\extras-images" upon image generation.

4. Upscaling Images with ReActor Extension

When you're swapping a particular face and wish to avoid searching for it every time you restart your Stable Diffusion web UI and drop it onto the ReActor Canvas, the ReActor Extension introduces a new feature. This feature allows you to save Face Models of faces you frequently use for face swapping. Let's delve into how this functionality works.

### Building & Saving Face Models
To do this, navigate to the "**Tools**" subtab within the ReActor Extension. Here, you can place a facial portrait onto the canvas, assign it a name, and click "**Build and Save**". This action will generate a .safetensors file containing the Face Model of your uploaded facial portrait. Below is an example of creating a Face Model for Angelina Jolie.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_face_models.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704461394/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_face_models_17bb1d3212.webp)

### Face Models Folder Location
Once you've constructed and saved your desired face model, locate the corresponding .safetensors file in the following directory: "stable-diffusion-webui\models\reactor\faces".

### Using Face Models for Easy Face Swapping
Now that you've generated your Face Model, let's explore how to employ it in upcoming face swaps. To utilize a face model, return to the "**Main**" subtab and opt for "**Face Model**". On the right side, you'll now find the option to select your built face model from the dropdown menu. In our example, you can observe that the "**Angelina**" Face Model has been included in the list. The web UI depiction below illustrates how it appears.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_face_model_dropdown.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704461783/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_face_model_dropdown_8fdfb9ebf0.webp)

**Note**: If your face model isn't visible, click the blue refresh icon next to the dropdown menu.

Once you've selected the right Face Model, pick a checkpoint, write a prompt, and witness the incredible results.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_face_model_example.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704462760/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_face_model_example_fb45c6850e.webp)

In the upcoming chapter we will have a look on how to Face Swap Multiple Faces. Let's dive in.

5. Saving & Using Face Models in ReActor

To initiate face swapping on an existing image, the primary requirement is to have an image readily available. This image can either be created through stable diffusion or sourced from the internet. With our chosen existing image ready, let's navigate to the "**img2img**" subtab within Stable Diffusion and proceed to drop it onto the canvas.

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-initial-image.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704899136/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_initial_image_922325ee4e.webp)

### Image2Image Settings (Denoising Strength)
Before delving into the ReActor Extension to discuss various face swap scenarios, let's adjust the denoising strength to 0. This ensures that, in addition to executing a face swap, we don't unintentionally modify the image. Now, let's proceed to open the ReActor Extension. Typically, upon opening the ReActor Extension, it is automatically enabled, indicated by the checked checkbox.

### ReActor Extension Settings
Instead of directly placing our chosen face portrait onto the "Singe Source Image", we can utilize the face models we created in the previous section. Click on "Face Model" and choose your preferred face model. In this instance, we'll opt for the "**Angelina**" face model we crafted earlier.

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-selecting-face-model.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704899482/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_selecting_face_model_5324eb249b.webp)

When selecting a face model, bear in mind that it essentially serves as your "**Source Image**". However, it is confined to a single face, and that face will consistently have a face number of 0.

### Target Images Face Numbers
Now, our aim is to exchange our Angelina Jolie source face with the target image, the one we placed on the image-to-image canvas. To grasp how we can seamlessly swap Angelina Jolie onto a specific face, let's delve into a deeper understanding of the concept of face numbers. The ReActor Extension will assign face numbers to the target image from left to right, commencing with face number 0. Therefore, our target image will have the following face numbers: 0,1,2.

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-face-number-target-image.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704899852/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_face_number_target_image_6fce220a58.webp)

Now that we are aware of the face numbers assigned to our target image, let's proceed to the "**target image**" section. Here, we can input the specific face number where we want to swap in Angelina Jolie's face. So let's say we want to swap the middle face of our target image with face number 1, we would input number 1 to swap the face. 

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-source-target-image-input-sections.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704899984/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_source_target_image_input_sections_f5738b0950.webp)

After clicking "**Generate**" you will observe that Angelina Jolie her face is swapped onto the middle face (target face number 1) of the target image.

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-swapped-face-example.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704900132/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_swapped_face_example_fd0fe372a8.webp)

And that's how easy it is to swap a face onto an existing image. You just have to know how to use the face numbers correctly.

### Face Swapping on Multiple Faces
Swapping a face onto an existing image is that simple! Just ensure you use the face numbers correctly. But there's more to it. What if you want to replace Angelina Jolie's face on multiple faces within the image? Well, you can achieve this by entering multiple face numbers in the "**target image**" section. Remember to include commas between the numbers without any spaces. 

For instance, if you want Angelina's face on both the left and right faces but not the middle, you would input something like: 0,2. Alternatively, you could input 2,0 since the order doesn't matter when working with a single source face. Once again, click generate and observe the result.

![reactor-face-swapping-crash-cours-face-swapping-on-existing-images-swapped-face-example-2.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704900302/reactor_face_swapping_crash_cours_face_swapping_on_existing_images_swapped_face_example_2_15a9a8d1ab.webp)


6. Face Swapping Existing Images

Mastering the ability to swap multiple faces simultaneously can be a valuable skill. By following this tutorial you might have come across a section within the extension where you have the "**Source Image**" and "**Target Image**" sections.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_multiple_face_swaps.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704466154/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_526a9abe14.webp)

Our goal is to clarify the sections by presenting a practical example.

### Explanation of Source Image
The source image is the picture placed on the ReActor Canvas. Typically, you'll upload a photo with a single face, and the default face number is 0 for that particular face. However, what if the image on the canvas features multiple faces?

Consider this scenario: I uploaded an image depicting the faces of two actresses. Now, I intend to assign numbers to each face, separated by commas. This enables you to specify which face you want the "**Target Image**" to adopt during the generation process. In this instance, I'll use: 0,1 when assigning face numbers.

_When assigning faces keep in mind that ReActor detects faces from left->right, top->bottom_

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_source_image.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704467815/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_source_image_8652f2cc2e.webp)

Assigned faces:

- Face number 0: Angelina Jolie
- Face number 1: Salma Hayek


### Target Image Explanation
In this section, input the face numbers (as designated in the "**Source Image**" section) that you desire for the output. For example, if you're creating an image featuring two women and you want the one on the right to have Angelina Jolie's face (Face Number 0) and the one on the left to have Salma Hayek's face (Face Number 1), ensure to correctly assign these numbers, like: 1,0 in the "**Target Image**" field.

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_target_image.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704468027/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_target_image_cb39388919.webp)

### Example Multiple Face Swapped Image

![stable-diffusion-reactor-course-face-swaps-reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_example.webp](https://res.cloudinary.com/db7mzrftq/image/upload/v1704468358/stable_diffusion_reactor_course_face_swaps_reactor_extension_crash_course_nextdiffusion_multiple_face_swaps_example_35d8a55b24.webp)

### Important Note
When using this approach, remember that if your Source Image has "X" faces, your output should have at least the same number of faces for it to work correctly. Adjust your prompt to make sure the output has the necessary minimum faces.


7. Face Swapping Multiple Faces

In this crash course, we'll swiftly guide you through the steps to download and leverage the ReActor extension within Stable Diffusion for achieving realistic face swaps.

OpenPose Crash Course (Stable Diffusion)

About this Chapter 👇

Chapters 👇

Share Your Experience ⭐

More Courses 🎓

ReActor Face Swap Crash Course (Stable Diffusion)

About

More Info

Explore AI Character Generation