Facefusion
Industry leading face manipulation platform
Last verified:
What is Facefusion?
FaceFusion is an industry-leading face manipulation platform that enables users to perform advanced face swapping, lip syncing, face enhancement, and various other face manipulation tasks on images and videos. It is an open-source Python-based application with a web-based UI and command-line interface, designed for technically skilled users who want professional-grade AI face manipulation capabilities.
Key features include 10+ processors such as face swapper (with Inswapper_128 and HyperSwap 256 models), deep swapper for alternative deepfake models, face enhancer (GFPGAN v1.4), lip syncer for audio-driven video, age modifier, expression restorer, face editor, frame enhancer, frame colorizer, background remover, face debugger, and voice extractor for separating vocals from music. The platform supports batch processing, job management with create/submit/run/retry functionality, pixel boost option for improved face quality, and deepfake webcam functionality.
FaceFusion is primarily for developers, AI researchers, content creators, video editors, and technical users who need sophisticated face manipulation tools. It requires terminal usage and technical skills for installation, though Windows and macOS installers are available for users uncomfortable with command-line. The platform has 25.6k stars on GitHub and supports Linux, macOS, and Windows with CUDA, CoreML, DirectML, MIGraphX, and OpenVINO accelerator support.
The tool processes images and videos locally on your machine, stores temporary resources in .facefusion folder in your OS temporary directory, and includes ethical safeguards blocking NSFW content. It supports multiple face detection models, face landmapping, face selection modes (one, many, reference), and face masking for precise control over manipulation areas.
Facefusion pricing
Pricing model: Freemium
FaceFusion is completely free and open-source with no paid tiers or premium versions. The core software is free forever. The project offers voluntary support through Buy Me a Coffee and Ko-fi platforms. They sell a Windows Installer and macOS Installer for users uncomfortable with terminal usage, but these installers contain the same open-source software without additional features - there is no FaceFusion Premium. All processors, features, and capabilities are included in the free version.
Facefusion pros
- Open-source and completely free to use
- 25.6k GitHub stars with active development
- 10+ specialized processors for different tasks
- Face swapper with Inswapper_128 and HyperSwap 256 models
- Lip syncer for audio-driven video animations
- Face enhancer with GFPGAN v1.4 for facial detail improvement
- Voice extractor separates vocals from music audio
- Job manager for batch processing multiple tasks
- Pixel boost option improves swapped face quality
- Supports CUDA, CoreML, DirectML, MIGraphX, OpenVINO
- Web-based UI plus command-line interface options
- Windows and macOS installers available for beginners
- Deepfake webcam functionality for real-time processing
- Age modifier and expression restorer processors
- Multiple face detection and landmapping models
- Frame enhancer and frame colorizer for video quality
- Background remover processor available
- Face editor for precise facial adjustments
Facefusion cons
- Installation requires technical skills and terminal usage
- Not recommended for beginners without CLI experience
- Minimum 8GB VRAM required, 12GB+ recommended
- No high-resolution inswapper models available
- NSFW content blocked due to ethical considerations
- Requires GPU for reasonable processing speeds
- Complex setup with conda environment configuration
- Multiple dependency installations can be challenging
Frequently asked questions about Facefusion
Are there plans to implement a high resolution inswapper?
InsightFace opted against releasing high-resolution inswapper models due to ethical considerations. The developers worked around this limitation by introducing the pixel boost option to improve swapped face quality within the constraints of available models.
How to improve the quality of swapped faces?
FaceFusion introduced the pixel boost option to work around model limitations and improve swapped face quality. Additionally, you can use the face enhancer processor with GFPGAN v1.4 model to enhance facial details like eyes and skin texture, with adjustable blend settings from 0-100.
What operating systems are supported?
FaceFusion supports the latest versions of Linux, macOS, and Windows. The platform is cross-platform and works on all three major operating systems with appropriate accelerator support for each.
How much VRAM should my GPU have?
FaceFusion considers 8GB VRAM as the bare minimum requirement. Everything from 12GB VRAM should work fine for smooth operation. More VRAM enables better performance and handling of higher resolution content.
Where can temporary resources be found?
Temporary resources are stored inside a .facefusion folder located in the temporary directory of your operating system. This is where FaceFusion caches intermediate processing data during face manipulation operations.
What causes the program to halt after 100% analysing?
The program halts after 100% analysing when it detects NSFW (Not Safe For Work) content. FaceFusion doesn't allow NSFW content due to ethical considerations, which is an intentional safeguard built into the platform.
Is there a paid premium version of FaceFusion?
No, there is no paid version or FaceFusion Premium. The author confirmed it is just a wrapper to an open-source library and is completely free. The Windows Installer and macOS Installer sold on Buy Me a Coffee contain the same open-source software without additional features - they are simply installer packages for users not comfortable using a terminal.
What processors are available in FaceFusion?
FaceFusion offers 10+ processors: age-modifier, expression-restorer, face-debugger, face-enhancer, face-editor, face-swapper, frame-colorizer, frame-enhancer, lip-syncer, background-remover, and deep-swapper. The default processor is face_swapper, but you can load multiple processors simultaneously using the --processors argument.
How do I install FaceFusion?
Installation requires 7 steps: 1) Prepare your platform, 2) Initialize conda and create environment with Python 3.12, 3) Install your accelerator (CUDA, CoreML, DirectML, MIGraphX, or OpenVINO), 4) Clone the repository from GitHub or Gitee, 5) Run install.py with appropriate onnxruntime option, 6) Reload your environment, 7) Run python facefusion.py run --open-browser. Windows and macOS installers are available for users uncomfortable with terminal.
Can I use FaceFusion for real-time face swapping?
Yes, FaceFusion includes Deepfake Webcam functionality that enables real-time face swapping. This feature allows you to use FaceFusion with webcam applications for live face manipulation during video calls or streaming.