FREE AI Deepfake: Control Expressions & Emotion | Image to Video with Live Portrait in Google Colab

Live Portrait is an advanced, open-source deepfake tool that allows you to animate static images by mapping a source video's expressions onto the photo. Developed by Quow, the same company behind Cling AI, Live Portrait captures complex facial expressions with impressive accuracy. For those of you interested in trying this powerful technology without the need for a graphics card, I will guide you through three online methods to use Live Portrait for free. Let's dive in!

Method 1: Using Hugging Face

Open the GitHub Repository: Access Live Portrait’s GitHub repo through the link provided.
Hugging Face Interface: Navigate to the Hugging Face page linked in the description where you can upload your source image and the driving video.
Aspect Ratio: Ensure that the aspect ratio of the video is 1:1.
Select or Upload Files: Use the example images and videos or upload your own. For demonstration, I used a Pixar-style image and a video with expressive movements.
Animate: Scroll down and click the “Animate” button. Wait a few seconds, and then you can download or play the animated video that replicates those expressions flawlessly.

The results are particularly impressive across various image styles such as black-and-white photos, realistic pictures, oil paintings, and even fictional statues.

Method 2: Using Replicate

Access the Replicate Page: Click the link in the description to reach the Replicate interface.
Upload Files: Change the default example files by clicking and uploading your own image. Enter the driving video URL if necessary.
Advanced Settings: Adjust settings like video frame load cap, selecting every nth frame, size scale ratio, lip and eye retargeting, etc. Default settings are generally sufficient.
Run: Hit the "Run" button and view the output. Note that Replicate does limit videos to a maximum of 5 seconds.

While Replicate offers more control than Hugging Face, its 5-second limitation may be a drawback for some users.

Method 3: Using Google Colab

Open Google Colab: Click the link in the description to navigate to the Google Colab page.
Setup GPU: Click 'Runtime' -> 'Change Runtime Type', and ensure the T4 GPU is selected. Then click 'Connect' at the top-right.
Run First Cell: Click the play icon, accept the warning by clicking 'Run Anyway'. Wait for the green check mark to confirm completion.
Upload Files: Go to the left panel, click on 'Files' to upload your image and video. Copy their paths and paste them in the corresponding fields in the second cell.
Run Second Cell: Click the play icon in the second cell. After it finishes, navigate to the live portrait animation folder to download your result.
Make Another Video: For additional videos, you only need to adjust the paths in the second cell and rerun it.

This method using Google Colab provides a seamless and efficient way to generate longer, high-quality animated videos.

These methods demonstrate the incredible potential of Live Portrait technology to animate static images with remarkable precision. Start experimenting and enjoy creating your very own animated portraits!

Keywords

Live Portrait
Deepfake tool
Hugging Face
Google Colab
Replicate
Image to video
Expressions mapping
Cling AI
Facial expressions
Static images animation

FAQ

Q1: What is Live Portrait? A1: Live Portrait is an advanced, open-source deepfake tool that maps a source video's expressions onto a static image to animate it.

Q2: How can I use Live Portrait without a graphics card? A2: You can use Live Portrait through three online methods: Hugging Face, Replicate, and Google Colab. Each method offers different capabilities and limitations.

Q3: What is Hugging Face and how does it work with Live Portrait? A3: Hugging Face provides an interface where you can upload an image and a source video, ensuring the video has a 1:1 aspect ratio. The tool then animates the image based on the video's expressions.

Q4: What are the limitations of using Replicate for Live Portrait? A4: Replicate offers more control over settings but limits the video output to a maximum of 5 seconds.

Q5: Can I run Live Portrait on Google Colab? A5: Yes, Google Colab is a powerful option that allows you to harness GPU capabilities. You upload your files, paste their paths, and run the process to generate animated videos.

Q6: What is the purpose of adjusting the paths in the second cell in Google Colab? A6: Adjusting the paths ensures that the correct image and video files are used for the animation process.