LivePortrait can generate dynamic videos with just 1 original image

Tutorial Introduction

This tutorial is a one-click run demo of the latest version of LivePortrait. The relevant environment and dependencies have been installed. You can experience it by cloning and starting it with one click.

LivePortrait is open sourced by the Kuaishou Keling Big Model Team. Its main functions include generating vivid animations from a single image, accurately controlling the movements of eyes and lips, processing seamless stitching of multiple portraits, supporting multi-style portraits, generating high-resolution animations, etc. This project uses an AI portrait animation generation framework based on an implicit key point framework. It can transfer the expressions and postures driving the video to static or dynamic portraits to generate highly expressive videos.

In the latest version, the model has added many interesting features such as precise portrait editing and video editing.

How to run

1. After cloning and starting the container, copy the generated API address to the browser to enter the Gradio interface inference experience

2. After opening the link, you will see the interface below. Upload pictures and videos respectively, click Animate and wait for a while to generate the video result.

• Image material: Upload a high-resolution photo of a front-facing portrait. Make sure the photo is sharp and evenly lit, with no overexposure or heavy shadows.
• Reference Video: Upload a reference video to drive the expressions and head movements of the portrait. The characters in the video can have conversations, sing, or make various expressions.

You can choose whether to enable the following 3 options in "Animation Options": relative motion, do crop, and paste-back

In the new version, you can also choose to focus more on the expressions or movements in the video, as well as parameters such as the smoothness of the movements.

Here are some examples:

3. Portrait video editing mode

In the interface below the new version, you can also choose whether to enter portrait video editing mode and make corresponding changes to the video.

4. Precise Portrait Editing mode

The next section of the interface is the Precise Portrait Editing mode. Users can choose to change the eyebrows, mouth size, wink, smile degree and other parameters to generate new pictures. At the same time, the position of the face can also be moved.

The effect is as shown in the figure

5. Retargeting Video Mode

At the bottom of the demo interface, you can also experience the Retargeting Video mode, which can adjust the size of the mouth opening of the character in the video. When the target lip-open ratio is adjusted to 0, you can see that the character's mouth is closed in the newly generated video.

Exchange and discussion

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓