Last updated 08-13-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
DeepFaceLab
DeepFaceLab trains and runs deepfake face swaps on your own GPU through a full Python pipeline you control locally. Extract faces from source and destination videos, train a neural network model, and merge the swapped face back into footage at resolutions up to 512 pixels and beyond.
Unlike cloud deepfake apps that hide the pipeline, DeepFaceLab gives you every stage, including S3FD face extraction, XSeg masking, and model architectures like SAEHD. That depth comes with a steep learning curve: the README warns there is no one-click fix and expects comfort with After Effects or DaVinci Resolve for finishing work.
VFX artists, YouTube creators, and researchers who need local, GPU-based face replacement are the core audience. The repository was archived read-only on November 13, 2024, but the code, Windows builds, and 19,300+ GitHub stars remain accessible. The related DeepFaceLive project handles real-time streaming swaps separately.
19,300+ GitHub stars with 928 forks and GPL-3.0 open-source license
Face replacement, de-aging, and full head swap workflows with native resolution training
S3FD face extractor and XSeg editor for precise face masking
Windows builds distributed via torrent and Mega.nz; Linux port available
Built on TensorFlow with CUDA and DirectX GPU acceleration
Related DeepFaceLive project adds real-time face swap for streaming and video calls
Full pipeline control from face extraction through model training to video merge
19,300+ GitHub stars and active community Discord for troubleshooting
Completely free under GPL-3.0 with no cloud dependency or usage limits
Supports face replacement, de-aging, and head swap in one toolkit
Repository archived read-only in November 2024 with no ongoing development
Requires NVIDIA GPU, CUDA setup, and significant time to learn the workflow
No one-click results; finishing work often needs After Effects or DaVinci Resolve
Is DeepFaceLab free to use?
Yes. DeepFaceLab is free and open source under the GPL-3.0 license. You download the code from GitHub or prebuilt Windows packages and run it locally on your own GPU hardware with no subscription fee.
Is DeepFaceLab still actively developed?
No. The DeepFaceLab GitHub repository was archived by the owner on November 13, 2024 and is now read-only. The code and releases remain downloadable, but no new commits or features are being added.
What can DeepFaceLab do?
DeepFaceLab supports face replacement in video, de-aging faces, and full head swaps. It trains custom neural network models from source and destination face sets, then merges the result back into the original footage.
What hardware does DeepFaceLab require?
DeepFaceLab requires an NVIDIA GPU with CUDA support for training and inference. The project uses TensorFlow and DirectX, and prebuilt Windows packages are the primary distribution method, with a community Linux port also available.
How does DeepFaceLab differ from DeepFaceLive?
DeepFaceLab focuses on offline video face swap training and merging with full pipeline control. DeepFaceLive, a related project by the same author, handles real-time face swapping for PC streaming and video calls.
Where can I get help with DeepFaceLab?
DeepFaceLab maintains an official Discord channel for English and Russian community support. The GitHub repository also has 530 open issues and documentation in the doc folder, though the repo is archived and read-only.