ComfyUI > Workflows > AnimateDiff + ControlNet + IPAdapter V1 | Flat Anime Style

AnimateDiff + ControlNet + IPAdapter V1 | Flat Anime Style

This ComfyUI workflow employs AnimateDiff, ControlNet (incorporating Depth, Softedge, and OpenPose), IPAdapter, Face Restore, Lora, among others, to convert original video content into a distinct Flat Anime Style. It streamlines the process, allowing for the creation of videos with a unique anime aesthetic effortlessly.

ComfyUI Vid2Vid (Anime) Workflow

Transform Video into Flat Anime Style Using AnimateDiff and ControlNet in ComfyUI

Want to run this workflow?

Fully operational workflows
No missing nodes or models
No manual setups required
Features stunning visuals

ComfyUI Vid2Vid (Anime) Examples

ComfyUI Vid2Vid (Anime) Description

1. ComfyUI Workflow: AnimateDiff + ControlNet + IPAdapter | Flat Anime Style

This ComfyUI workflow utilizes AnimateDiff, ControlNet featuring on Depth, Softedge, etc., IPAdapter, and FaceRestore to transform original video content into a distinctive Flat Anime Style. After obtaining the result, you can activate the upscale nodes to enhance your video's resolution.

2. Overview of AnimateDiff

Please check out the details on

3. Overview of ControlNet

Please check out the details on

4. How to Use Face Restore

"FaceRestore" in ComfyUI is a custom extension designed for restoring faces in images. It leverages the capabilities of the CodeFormer model to enhance image fidelity. Here are the detailed explanations.

4.1. Input of “Face Restore CF With Model” node

facerestore_model: Specify the face restoration model to use. This is essential for defining the algorithm that will be applied to enhance the faces in your images.

image: This is the input image that contains faces you wish to restore. The node will process this image and apply face restoration on detected faces.

facedetection: Choose the face detection model from the following options. This model is responsible for identifying and cropping faces from the input image: Each of these options has its strengths, with some being more accurate while others are faster or lighter in terms of computational resources required.

retinaface_resnet50
retinaface_mobile0.25
YOLOv5l
YOLOv5n

codeformer_fidelity (FLOAT): A critical parameter that allows you to adjust the fidelity of the CodeFormer model. This setting determines the balance between restoring the face with high fidelity to the original and enhancing the image. A higher value might retain more original features, while a lower value may result in a more 'idealized' restoration.

4.2. Output of “Face Restore CF With Model” node

IMAGE: The output is the processed image where the faces have been restored. This image is the result of the face restoration process, showcasing enhanced clarity, details, and overall improved visual quality of the faces detected in the input image.