How To install models in invoke ai And using ip adapter plus Face

28 Sept 202311:54

TLDRThe video script offers a step-by-step guide on installing and using IP adapters from, focusing on the face and image processing capabilities. It emphasizes the importance of having the correct software version and provides instructions on importing models via the model manager. The tutorial demonstrates how to load images, adjust settings for image-to-image generation, and fine-tune the output using various tools and parameters. The end goal is to achieve a high level of accuracy in facial feature adaptation while maintaining the original image's context, showcasing the versatility and potential of the IP adapter technology.


  • πŸ” The process begins by visiting to select desired models from a list categorized under Concepts, Styles, and Tools.
  • πŸ“Œ It's crucial to check the model versions and download the correct encoder to ensure compatibility with the existing setup.
  • 🎯 The IP adapter plus face model is highlighted as a preferred choice for its ability to generate images based on facial input.
  • πŸ› οΈ To add a model, users should navigate to the model manager, which is symbolized by a cube icon, and then proceed to import models.
  • πŸ“‚ The import process involves pasting the copied model link without an HTTP prefix, as the system does not accept it.
  • πŸ–₯️ The terminal can be used to monitor the progress of model installation, providing more transparency than the UI.
  • πŸ”„ Syncing models in the settings can help to ensure all models are up-to-date, though it's unclear if this applies to all, including the IP adapter.
  • πŸ–ΌοΈ Users are guided to load images with specific dimensions, maintaining aspect ratios for consistency, and utilize the asset browser for easy access.
  • 🎨 The script discusses the use of image-to-image mode with prompts, and the importance of adjusting settings like denoise strength for desired outcomes.
  • πŸ”§ Tweaking blend modes and other parameters allows users to fine-tune the generated images to closely match the input, including skin tone and background elements.
  • βœ… The final step involves testing different settings and adjustments to achieve the most accurate and satisfactory results before finalizing the image.

Q & A

  • What is the first step in installing models from

    -The first step is to visit the website and navigate to the 'models' section at the top of the page.

  • How can you determine the required version for a model?

    -By hovering your cursor over the yellow icon (circle) next to the model, it will display the necessary version information.

  • What is the IP adapter plus face model used for?

    -The IP adapter plus face model is used to input a face which is then passed as a condition for the image to generate a similar face.

  • How do you add a model in the models manager?

    -You add a model by copying the model information from the website and pasting it into the 'import models' section in the models manager.

  • Why is it important to sync models in the settings?

    -Syncing models in the settings ensures that all models are updated and consistent across different uses, although it's unclear if it's necessary for all types of models like the IP adapter.

  • What are the image dimensions used in the example?

    -The image dimensions used in the example are 384 by 512 and 512 by 768, with proportions maintained so that when divided by two, the dimensions match.

  • How does the denoise strength setting affect the output?

    -The denoise strength setting affects how closely the output resembles the input image; a higher setting will result in less similarity, while a lower setting will try to maintain more of the original image's details.

  • What is the purpose of the blend mode?

    -The blend mode allows for adjusting how much of the original image is retained in the final output, preventing unwanted changes to certain elements like the jacket or background.

  • How can you make the face in the image more accurate?

    -To make the face more accurate, you can adjust the blend mode settings, tweak the denoise strength, and use the image action to send the image to a unified canvas for further adjustments.

  • What are the key hotkeys mentioned for hiding the mass and the box?

    -The key hotkeys mentioned are Shift + C to hide the mass and Shift + H to hide the box.

  • What is the final step to finalize the IP adapter face model adjustments?

    -The final step is to click the checkmark to confirm the adjustments, which then allows for the model to generate the final output with the desired face and retained background elements.



πŸ“¦ Installing Models from

This paragraph provides a step-by-step guide on how to install models from the website It starts by navigating to the site and selecting desired models, emphasizing the importance of having the correct version to ensure compatibility. The speaker mentions their preference for IP adapters, explains the process of adding a model by copying and pasting the model link into the model manager's import section, and highlights the absence of HTTP in the model link. The paragraph also touches on the installation process, including monitoring progress through the terminal for a more detailed status update. Lastly, it mentions syncing models in settings for convenience.


πŸ–ΌοΈ Using the IP Adapter with Images

The second paragraph delves into the application of the IP Adapter with images. It discusses the process of loading images into the system, noting the importance of maintaining aspect ratios for consistency. The speaker describes using two different images with a prompt and adjusting settings to achieve desired variations. The focus is on manipulating denoise strings to control the level of variation in the generated images. The paragraph also explains the use of blend mode to achieve a closer match to the original image while maintaining certain features like the jacket and background. The speaker provides practical tips on tweaking settings for optimal results.


🎨 Fine-Tuning the IP Adapter for Accuracy

This paragraph centers on refining the IP Adapter's output for greater accuracy. It begins with an assessment of the initial results, noting the need for adjustments to better match the original image. The speaker details the process of tweaking the blend mode and denoise strength to improve the facial features while preserving the original image's elements such as the jacket and background. The paragraph also introduces the concept of sending the image to a unified canvas for further refinements and suggests making final adjustments to achieve the desired outcome. The speaker's approach to fine-tuning demonstrates a methodical and iterative process, emphasizing the importance of patience and precision in achieving accurate and satisfactory results with the IP Adapter.




The term '' refers to a website that hosts a variety of models, which are likely software tools or algorithms used for specific tasks. In the context of the video, it is the source from which the user is instructed to install models. The video demonstrates navigating this site to select and install desired models, indicating its importance as a platform for accessing these tools.

πŸ’‘IP adapter plus face

The 'IP adapter plus face' is a specific model mentioned in the script that seems to be designed for processing images, particularly focusing on faces. It takes an input face image and generates a similar face, which can be used in various applications. The video provides a tutorial on how to install and use this model, highlighting its utility in image manipulation and generation.


An 'encoder' in this context appears to be a software component necessary for the functioning of the models. The script emphasizes the importance of having the correct version of the encoder, suggesting that it plays a crucial role in the compatibility and effectiveness of the installed models. The encoder could be responsible for encoding data in a way that the models can process.

πŸ’‘Model Manager

The 'Model Manager' is a tool or interface within the software that allows users to manage their installed models. It is where users can import, organize, and possibly configure the models they have added. The video describes using the Model Manager to import the 'IP adapter plus face' model, indicating its role in the workflow of setting up and using the models.


HyperText Transfer Protocol (HTTP) is a system for transferring data over the internet. In the video, it is mentioned in the context of importing models, where the user is accustomed to using HTTP for websites but is informed that it is not required for importing models into the system. This suggests that the process of model importation in this software is different from typical web browsing or data transfer.


The 'Terminal' is a command-line interface for interacting with the operating system or software. In the video, the terminal is used to monitor the progress of model installation, providing more detailed information than the user interface. This suggests that the terminal offers a deeper level of control and insight into the operations happening within the software.


In the context of the video, 'sync' refers to the process of synchronizing or updating models to ensure that they are consistent across different devices or platforms. The user mentions syncing models in their settings, which implies that this action updates or backs up the models to maintain uniformity and up-to-date status.

πŸ’‘Asset Browser

The 'Asset Browser' is a feature within the software that allows users to browse and manage their assets, which could include images, models, or other files used in their projects. The video describes using the asset browser to load images for processing with the 'IP adapter plus face' model.


The 'Denoiser' is a tool or setting within the software that reduces noise or unwanted variations in the output of the models. In the video, the user adjusts the denoiser strength to control the level of detail and similarity between the input and output images, affecting the quality and accuracy of the generated faces.

πŸ’‘Blend Mode

The 'Blend Mode' is a feature that allows for the combination of different layers or elements in an image. It determines how the colors and properties of one layer interact with those beneath it. In the video, the blend mode is adjusted to achieve a more accurate and harmonious integration of the generated face with the existing image.

πŸ’‘Unified Canvas

The 'Unified Canvas' seems to be a workspace or platform within the software where users can manipulate and adjust their images or projects. It offers tools for further refinement and editing after the initial model processing. The video describes sending the image to the unified canvas for additional adjustments.


The process of installing models from is described, providing users with a step-by-step guide on how to access and utilize various models.

The importance of having the correct version of the software is emphasized to ensure compatibility with the models, highlighting the need to download the appropriate encoder.

The IP adapter plus face model is introduced as a preferred choice, with its functionality to input a face and generate a similar one in the output image.

The user interface of the models manager is explained, including the import models feature and the absence of HTTP in the model import process.

The process of syncing models through settings is mentioned, suggesting it as a potential best practice for users to keep their models updated and consistent.

The use of image to image functionality is detailed, explaining how to load images into the asset browser and apply settings for optimal results.

The significance of aspect ratio in image selection is discussed, with a preference for maintaining consistent proportions for better outcomes.

The utilization of prompts and settings in the image to image process is highlighted, showing how they can be adjusted to refine the final output.

The impact of denoise strength on the variation of generated images is explained, with guidance on finding the right balance for desired results.

The blend mode setting is introduced, with its role in achieving a closer match between the input and output images.

The process of fine-tuning the model's output by adjusting settings such as denoise strength and blend mode is described, emphasizing the iterative nature of achieving accurate results.

The use of the unified canvas for further adjustments and refinements of the generated images is mentioned, showcasing additional options for enhancing the final product.

The transcript provides practical tips on achieving a good skin tone match and maintaining desired elements in the background of the generated images.

The transcript concludes with a summary of the IP adapter for the face process, highlighting its effectiveness and potential applications.