ComfyUI Tutorial: Zero-based mastery of IPAdapter local deployment, face replacement, style imitation rely on it!

IPAdapter is a very powerful plugin with three main features:

First,face replacement.

IPAdapter can accurately recognize the facial features of the people in the reference map and introduce the features to the resultant map to achieve the face replacement effect.

347fb95fj00szmtw300z7d000sf00igm

The second is material migration.

In addition to recognizing the facial features of a person, IPAdapter is also able to recognize the material of the reference image and apply it to the resultant image.

bb620f31j00szmtw300q0d000u000igm

Finally, there is style migration.

IPAdapter, in addition to the above two functions, recognizes the style of the reference graph and migrates the results to the result graph.

2784058dj00szmtw301jqd000u000j4m

IPAdapter Local Deployment

I. Plug-in Installation

If you are using the Autumn Leaves Integration Pack, the IPAdapter_plus plugin is already installed by default, so you don't need to install the plugin again.

83f79411j00szmtw1001ad000tl00adm

If you don't have IPAdapter plugin installed locally, you can install it by searching the manager and installing the first "IPAdapter_plus".

6e61a8fdj00szmtw1002fd000u0008xm

You can also install it via the github link

https://github.com/cubiq/ComfyUI_IPAdapter_plus.git

d4106ba9j00szmtw20052d000u000g6m

For more plugin installation methods, see the article:

AI Painting ComfyUI Getting Started Tutorial, ComfyUI Workflow Custom Node Installation

2. Download the model

Ipadapter's plugin is better to install, the hard part is the model installation, you must be more careful when installing.

Still open the Ipadapter's github page and scroll down to see the download link for its models, which are divided into 4 main categories.

(https://github.com/cubiq/ComfyUI_IPAdapter_plus.git)

1、Download the visual model, place the path:/ComfyUI/models/clip_vision

(Note: the name of the downloaded model needs to be renamed to be the same as the name in blue in the picture below)

bd6aa5adp00szmtw1000pd000jd0040m

2、Download the basic model, facial model, etc., model placement path: /ComfyUI/models/ipadapter, if there is no ipadapter this folder, create a new one yourself.

(Discarded models may not be downloaded)

0bf6a91fj00szmtw1003dd000lv00bnm

3、Download the basic faceid model, etc., model placement path: /ComfyUI/models/ipadapter

(Discarded models may not be downloaded)

aef608a7j00szmtw1002sd000ki009xm

4、Download the lora model, place the path: /ComfyUI/models/loras

3748a514j00szmtw1001rd000rr005zm

III. Simple use of workflows

Come to the comfyui_ipadapter_plus plugin directory, find the examples folder, this folder has a lot of sample workflows, load a default workflow.

Key point: the presets in the IPAdapter loader actually represent models with different weights, here we use light, standard, medium, high intensity, etc. to help you understand better.

4ab5f59ej00szmtw1000sd000gh008tm

Change the preset in "IPAdapter Loader" to medium intensity and add a prompt:

Cue word: closeup of a fierce warrior woman wearing a full armor at the end of a battle, high quality, detailed)

Click Generate Image.

The generated images do not follow the cue words, but the style, color, and composition of the generated images are very similar to the reference image.

6a002b6bj00szmtw200bpd000u000eum

At this point, change the weight type in the "Apply IPAdapter" node to "Prompts are more important".

0987d859j00szmtw1000rd000c4009rm

The images generated now follow the cue word more closely, while the style also takes into account the characteristics of the reference image.

73a8937ej00szmtw300nyd000u000qtm

We then change the weight type to style conversion, which is now the style conversion effect on the reference image.

d308cf35j00szmtw300fzd000u000jsm

Change the "IPAdapter Loader" preset to Portrait, and change the "Apply IPAdapter" weight type to Cue word is more important, and then load a frontal portrait, and the generated image is following the cue word for the character scene, but the face is Refer to the picture. This realizes face replacement.

e618d60cj00szmtw200drd000u000elm

These are the main points of today's post, which focuses on the installation of the IPAdapter plugin and model, as well as the simple usage of workflows.

statement:The content of the source of public various media platforms, if the inclusion of the content violates your rights and interests, please contact the mailbox, this site will be the first time to deal with.
TutorialEncyclopedia

How to generate new images with high similarity from existing images?AI one-click backpropagation of image cue words to replicate similar images

2025-7-21 9:55:31

TutorialEncyclopedia

Use AI to give popular film and television drama "Erchuang dubbing" music cover, AI cover any song and voice

2025-7-22 9:31:43

Zenium
Search