Deep-Live-Cam 2.1.6: Real-Time Face Swap with Just One Photo

Imagine swapping your face with a celebrity in a live video call, or turning yourself into a movie character in real time — all with a single photo. That is exactly what Deep-Live-Cam 2.1.6 lets you do. This open-source project has taken the internet by storm, and in this guide I’ll show you what it is, how to install it, and how to use it responsibly.

What is Deep-Live-Cam?

Deep-Live-Cam is a real-time face swap and video deepfake tool. You give it a single image of a person's face, and it replaces that face in a webcam stream or pre-recorded video — live, with no model training involved. It runs on CPU alone, but for smooth performance you’ll want an NVIDIA GPU, Apple Silicon, or even an Intel processor with the right execution provider.

Ethical use is a must

Before the fun, the serious stuff. The project has built-in checks that block nudity, graphic content, and sensitive material like war footage. The developers also insist that if you use a real person’s face, you get consent and label output as a deepfake. You are responsible for what you do with this software.

What can you do with it?

Manual installation steps

Deep-Live-Cam isn't an install-and-go app, but it isn't too difficult either. You'll need Python (3.11 to 3.14, with 3.14 recommended), pip, git, ffmpeg, and on Windows the Visual Studio 2022 runtimes. Start by cloning the repository:

Next, download two model files — gfpgan-1024.onnx and inswapper_128_fp16.onnx — from the Hugging Face model repository and put them in the models folder. Then create a virtual environment and install dependencies:

Mac users with Apple Silicon (M1 through M5) should first install Python 3.14 and python-tk@3.14 via Homebrew, then run the same venv commands. If tkinter goes missing, reinstall python-tk.

How to use Deep-Live-Cam

Once everything is installed, launch the GUI by running python run.py. Select a source face image, choose your camera, and click Live. After about 10 to 30 seconds of loading, a preview window appears. That feed can be captured with OBS for streaming. To swap to a new face, just select another source image.

There is also an image and video mode: pick a target image or video, press Start, and the output is saved in a new folder. The CLI mode is activated when you call the script with a source argument, and advanced options include --many-faces, --mouth-mask, and --live-resizable.

GPU acceleration options

CPU works but can be choppy. NVIDIA users can install CUDA Toolkit 12.8 and cuDNN, replace onnxruntime with onnxruntime-gpu, and launch with --execution-provider cuda. Mac users can use the built-in CoreML provider. Windows users can try DirectML for AMD, and Intel users can use OpenVINO. Each has its own setup notes in the README, so check the exact package versions there.

Want a zero-setup version?

If you don't want to juggle Python versions and model files, there is an official pre-built distribution called Deep-Live-Cam Ultimate. It includes optimized builds for Windows, Mac Silicon, CPU, NVIDIA, and AMD, plus 30+ extra features and priority support. It costs money, but it saves time. Just remember: the only official sources are the GitHub repository and deeplivecam.net, so avoid sketchy third-party downloads.

The bottom line

Deep-Live-Cam is an impressive leap in AI-driven media tools. It is powerful, easy to use once installed, and genuinely fun to experiment with. Just stay on the right side of consent and content rules. Now go give yourself a movie star face — responsibly.