DeepFaceLive

Overview

DeepFaceLive is a free, open-source tool that replaces your face in real-time during live video streams and calls. Rather than swapping faces in recorded footage after the fact, it processes webcam input frame-by-frame using GPU-accelerated AI, showing the processed result in a DeepFaceLive output window. For video calls or streaming, capture that window in OBS and use OBS Virtual Camera as the camera input for Zoom, Skype, Discord, or similar apps.

The author archived the repository on Nov 13, 2024. Because it is read-only, users should not expect official commits, releases, or compatibility fixes unless the maintainer reopens development. Existing releases remain downloadable, but compatibility with current drivers, operating systems, and third-party applications is not guaranteed.

DeepFaceLive remains a reference implementation for users who can work within its documented Windows and GPU constraints, but its archived state is a decisive limitation for new adoption.

Key Features

• Face Swap (DFM — Deep Face Model) — Swap faces using trained DFM models. The official README lists about 30 ready-to-use public face models and notes that most listed personas are fictional, with Keanu Reeves as an exception; users can also train custom DFM models using DeepFaceLab if they need a specific face not in the library.
• Face Swap (InsightFace) — Swap faces using a single uploaded photo rather than a trained model. Faster to set up than DFM; best suited for casual use cases where a pre-trained model is unavailable.
• Face Animator — Drive a static face image (photo or illustration) using webcam or video input; the official README describes ~25 fps in a hardware-dependent example using a 35 TFLOPS GPU. Useful for virtual avatar presentations and animated character streaming — an alternative to dedicated AI avatar generator tools for live use cases.
• OBS Studio Integration — DeepFaceLive outputs a processed preview window that OBS can capture as a Window Capture source; OBS Virtual Camera can then expose the result to streaming or video-call apps without platform-level changes.
• Multi-GPU Support — Leverages DirectX12 for rendering, with support for multiple GPU configurations for users with higher performance requirements.
• Multiple Face Detectors — Ships with several face detection algorithms selectable per session, allowing users to optimize for accuracy versus speed depending on their hardware.

How to Get Started

DeepFaceLive requires a compatible GPU, Windows 10, and a video driver update before the application runs correctly. The setup process is straightforward but not zero-configuration:

1. Update GPU drivers — Install the latest DirectX12-compatible driver for your graphics card (NVIDIA or AMD). The official README lists any DirectX12-compatible graphics card as the minimum and recommends RTX 2070+ or Radeon RX 5700 XT+.
2. Download the release — Go to the GitHub releases page and download the latest self-extracting archive. No Python installation or dependency management is needed — the release is portable and self-contained.
3. Extract and launch — Run the extracted executable. DeepFaceLive opens a multi-panel interface where you configure the video source, face detector, swap mode (DFM or InsightFace), and output settings.
4. Load a face model — For DFM mode, download a pre-trained .dfm model from the community model library and load it into the application. For InsightFace mode, upload a clear frontal photo of the target face.
5. Configure virtual camera — Set DeepFaceLive's Stream output to a window/merged-frame preview. In OBS, add the DeepFaceLive output window as a Window Capture source, start OBS Virtual Camera, and select OBS Virtual Camera in your video-call or streaming app.

System RAM requirement is 4GB minimum with a 32GB+ Windows paging file recommended for stable extended sessions.

How It Compares

The durable comparison is maintenance and deployment model. DeepFaceLive is a free, local, Windows-focused project whose upstream repository is archived. Anyone choosing an alternative should verify that project's current repository status, supported platforms, licensing, hardware requirements, and pricing on its own official sources; this pilot does not publish a cross-vendor matrix.

Pricing & Plans

DeepFaceLive is completely free. No subscription, no credits, no paid tier. The GPL-3.0 license allows free use, modification, and distribution for personal and commercial purposes, subject to the license's copyleft terms (derivative works must also be GPL-3.0).

There are no cloud processing costs — all inference runs locally on your GPU. The only cost is the hardware itself.

Community & Ecosystem

The original repository, issues, releases, and discussions remain available as historical references. Community support and third-party model libraries are not official maintenance channels and were not verified for this update; treat downloads from them as untrusted until reviewed.

Best For

• Streamers and content creators who want real-time face replacement in OBS-based workflows on Windows hardware with an RTX 2070+ or equivalent GPU
• Developers and researchers who need an open-source, locally-runnable face swap baseline for experimentation or building derivative tools
• Users who want zero ongoing cost and are comfortable with manual GPU driver configuration and parameter tuning
• AI enthusiasts exploring the best AI face swap tools who prefer locally-executed, privacy-preserving inference over cloud-based services
• Legacy users already running DeepFaceLive who don't need updates and want to continue using a stable, known setup

Avoid If

• You need active upstream maintenance, security patches, or guaranteed compatibility with current drivers; the official repository has been archived and read-only since November 2024.
• You need a supported macOS, Linux, or hosted workflow rather than a Windows-focused local setup.
• You cannot accept the consent, impersonation, and platform-policy risks that accompany real-time face replacement.

Sources & Verification

• Last verified: 2026-08-20
• Evaluation scope: Repository lifecycle, supported setup described by the maintainer, licensing, releases, and operational constraints; output quality was not benchmarked.
• Method: Source-based research from the official product materials below. No hands-on testing was performed for this update.
• Official sources:
  • DeepFaceLive repository
  • DeepFaceLive releases

FAQ