explaingit

hacksider/deep-live-cam

🔥 Hot★ 93,116PythonAudience · vibe coderComplexity · 3/5ActiveLicenseSetup · hard

TLDR

AI tool that swaps a face onto live video or pre-recorded footage in real time using a single reference photo.

Mindmap

mindmap
  root((repo))
    What it does
      Face swap live video
      Face swap recorded video
      Mouth-mask mode
      Multi-face mapping
    How it works
      Single reference photo
      Real-time processing
      Pre-trained models
      GPU acceleration optional
    Use cases
      Content creation
      Character animation
      Movie watching
      Live streaming
    Safety and ethics
      Built-in content filter
      Consent requirement
      Deepfake labeling
      Local law compliance
    Tech stack
      Python 3.11
      FFmpeg
      Pre-trained AI models

Things people build with this

USE CASE 1

Create deepfake videos for entertainment, memes, or content creation by swapping faces in live streams or recorded footage.

USE CASE 2

Animate characters or avatars by mapping a performer's face onto a digital character in real time.

USE CASE 3

Watch movies or shows with a different actor's face substituted onto the original speaker.

USE CASE 4

Run interactive live shows or streams where your face appears on a different body or character.

Tech stack

PythonFFmpegDeep learning models

Getting it running

Difficulty · hard Time to first run · 1h+

Requires downloading large deep learning models and GPU/CUDA setup for real-time performance.

Use it freely, but if you run it as a network service, you must release your changes to users. Strongest copyleft for SaaS.

In plain English

Deep-Live-Cam is an AI-powered tool that swaps a face onto a live video feed in real time, using only a single source photo as the reference. The README's three-click summary captures the user experience: pick a face image, choose which camera to use, and press "live", from then on, whatever the webcam captures appears with that other face superimposed on the speaker. The same approach also works on pre-recorded videos to produce a video deepfake without needing many training images. The program comes with a built-in safety check: it refuses to process media the developers consider inappropriate, including nudity, graphic content, and sensitive material such as war footage. The README dedicates a long disclaimer to ethical use, asking users to obtain consent when swapping a real person's face, to clearly label outputs as deepfakes, and to follow local laws. The project is written in Python (the README recommends version 3.11) and depends on tools such as ffmpeg, plus pre-trained model files that the user downloads separately. There are setup paths for Windows, Linux, and Apple Silicon Macs, with optional GPU acceleration. A pre-built quickstart download is offered for non-technical users. Beyond face swapping the README highlights features such as a mouth-mask mode that keeps the original speaker's mouth movement, mapping different faces onto multiple subjects in the same scene, watching movies with a substituted face, running live shows, making memes, and using it on chat sites. Someone might use it for content creation, character animation, or experiments with AI-generated media.

Copy-paste prompts

Prompt 1
How do I set up Deep-Live-Cam on my Windows machine and start swapping faces in a live webcam feed?
Prompt 2
Show me how to use the mouth-mask mode to keep the original speaker's mouth movements while swapping their face.
Prompt 3
How can I map multiple different faces onto different people in the same video scene?
Prompt 4
What are the ethical guidelines and safety checks I need to follow when using Deep-Live-Cam to create deepfakes?
Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.