First Order Motion Model (Damedane, Mayiyahei) Tutorial

Due to Bilibili's review policies, uploaders are not allowed to upload tutorial videos related to "AI face-swapping". However, the review of output content is relatively lenient, depending on specific content. Therefore, a text version tutorial for creating damedane online is provided on this site for reference. Through this tutorial you can:

• Generate damedane, Mayiyahei, unravel and other AI face-swapping videos for free
• Create entirely online without downloading software, only need to provide materials
• Learn the most basic usage of JupyterLab

Update Log

• Added "Mayiyahei" video material
• Fixed audio-video sync issues in output
• Optimized video output

First, visit the following link to register:

https://app.hyper.ai/console/signup

After registration, you will be automatically logged in. Then enter the console homepage and click on Public Tutorials in the left sidebar:

Then select damedane from the list on the right:

After entering the tutorial page, there is a clone button in the upper right corner. Click clone to clone this tutorial to your own account:

Then choose any container name. For computing power, you can select either vGPU or CPU. vGPU will be faster, but CPU has more free quota, though relatively slower. The parameters below don't need to be changed. Just use the default PyTorch image, then directly click Review and Execute:

Then wait a moment. When the running status changes to "Running", you can click "Open Jupyter Editor" at the bottom of the page:

In the Jupyter editor, double-click damedane.ipynb on the left side to open the corresponding file:

In the file list on the left, you can see two files: import.mp4 and import.png, where import.mp4 is the original damedane video file, and unravel.mp4 is the Unravel source file.

This tutorial runs unravel by default. That is, if you execute it without any modifications, the face-swapping video created will be the unravel pain video. If we need to create a damedane face-swapping video, we need to:

• Modify the source_image_name variable for the input image in the code
• Modify the source_video_name variable for the input video in the code
• Execute the task

Next, let's use creating damedane as an example. First, prepare the image you want to use. Note that the image should not be too large, preferably 256x256 pixels, then upload it directly to the file list on the left by dragging and dropping. Or you can also upload by clicking "Upload Files" above.

Then in the code block on the right, scroll down to the code block marked [20]:. Replace import.png with the filename of the image you just uploaded:

Then click "Run - Run All Cells" in the menu bar above to execute all code blocks. Then wait 3-5 minutes

Tip: Code blocks that are executing or waiting to be executed will display a [*]: symbol on the left side. Additionally, completed code blocks will show the corresponding execution time below:

After execution is complete, you can see the preview, and then follow the instructions in the last code block to proceed: