LivePortrait: No-GPU Cloud Tutorial - RunPod, MassedCompute & Free Kaggle Account - Animate Images
With V3 update video to video added. Are you interested in using LivePortrait, the open-source zero-shot image-to-animation application, but lack a powerful GPU, you are Mac user or prefer to use it in the cloud? If so, this tutorial is exactly what you need. I will guide you through the process of installing and using the LivePortrait application with just one click on #MassedCompute, #RunPod, and even on a free #Kaggle account. After following this tutorial, you'll find running LivePortrait on cloud services as straightforward as running it on your own computer. LivePortrait is the latest state-of-the-art static image to talking animation generator, outperforming even paid services in both speed and quality.
🔗 LivePortrait Installers Scripts ⤵️
▶️ / 119254105
🔗 Windows Tutorial - Watch To Learn How To Use ⤵️
▶️ • Animate Static Photos into Talking Videos ...
🔗 Official LivePortrait GitHub Repository ⤵️
▶️ https://github.com/KwaiVGI/LivePortrait
🔗 SECourses Discord Channel to Get Full Support ⤵️
▶️ / discord
🔗 Paper of LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control ⤵️
▶️ https://arxiv.org/pdf/2407.03168
🔗 Upload / download big files / models on cloud via Hugging Face tutorial ⤵️
▶️ • How to Download (wget) Models from CivitAI...
🔗 How to use permanent storage system of RunPod (storage network volume) ⤵️
▶️ • How To Use Storage Network Volume In RunPo...
🔗 Massive RunPod tutorial (shows runpodctl) ⤵️
▶️ • Ultimate RunPod Tutorial For Stable Diffus...
0:00 Introduction to the state-of-the-art image to animation open source application LivePortrait cloud tutorial
2:26 How to install and use LivePortrait on MassedCompute with amazing discount coupon code
4:28 How to enter our special Massed Compute coupon to get 50% discount
4:50 How to setup ThinLinc client to connect and use Massed Compute virtual machine
5:33 How to setup synchronization folder of ThinLinc client to transfer files between your computer and MassedCompute
6:20 How to transfer installer files into Massed Compute sync folder
6:39 How to connect initialized Massed Compute virtual machine and install LivePortrait app
9:22 How to start and use LivePortrait application on MassedCompute after installation has been completed
10:20 How to start second instance of LivePortrait on the second GPU on Massed Compute
12:20 Where the generated animation videos are saved and how we can download all of them to our computer
13:23 How to install LivePortrait on RunPod cloud service
14:54 Which template of RunPod you need to use
15:20 How to setup RunPod proxy access ports
16:21 How to upload installer files into JupyterLab interface of RunPod and start installation process
17:07 How to start LivePortrait app on RunPod after installation has been completed
17:17 How to start LivePortrait on the second GPU as second instance
17:31 How to connect LivePortrait from proxy connection of RunPod
17:55 Animating first image on the RunPod instance with 73 seconds driving video
18:27 How much time animating 73 seconds video takes (speed of the app is mind blowing)
18:41 How to understand input upload error and example case
19:17 How to 1-click download all the generated animations on RunPod
20:28 How to see and follow progress of the generating animations
21:07 How to install and start using LivePortrait for free on a free Kaggle account and speed is amazing
24:10 Generating first animation on the Kaggle after installed and started the LivePortrait app
24:22 Wait input images and videos to be uploaded fully or you will get error shown here
24:35 How to watch the status of the animation and follow the progress on Kaggle
24:45 How much GPU, CPU, Ram and VRAM is being used and the speed of animation process of LivePortrait app on Kaggle
25:05 How to download all of the generated animations on Kaggle with 1-click
26:12 How to restart LivePortrait app on Kaggle without reinstalling
26:36 How to join SECourses Discord channel to chat with us and get help
LivePortrait paper presents LivePortrait, an innovative framework for animating static portrait images into realistic and expressive videos. The authors focus on achieving high inference efficiency and precise controllability while maintaining high-quality results.
The proposed method builds upon and extends the implicit-keypoint-based framework, balancing computational efficiency and controllability.
Key improvements include:
Enhanced generation quality and generalization:
Scaled up training data to 69 million high-quality frames
Adopted a mixed image-video training strategy
Upgraded network architecture
Designed better motion transformation and optimization objectives
Improved controllability:
Introduced stitching and retargeting modules using small MLPs
Enabled precise control over eyes and lip movements
Allowed seamless animation of multi-person portraits
The model consists of two training stages:
Stage I: Base Model Training