lostwarmonger/deep-live-cam-portable

---

license: mit

Deep-Live-Cam Portable Environment (Windows & NVIDIA)

This is a complete, pre-packaged Conda environment for the Deep-Live-Cam face-swapping tool. It saves you from manually installing CUDA, PyTorch, or hunting for DLL files.

⚠️ Important Requirements (Must Read)

1. Operating System: Windows 10 or 11 (Only Windows is supported) .
2. Hardware: NVIDIA GPU (RTX 20, 30, or 40 series).
3. Prerequisite: You must have Miniconda (or Anaconda) installed on your system.

🚀 Installation & Usage (3 Steps)

Step 1: Download the File Download the deeplivecam.tar.gz file from this repository.

Step 2: Extract to Conda Open a terminal (Anaconda Prompt) and run:

# Create the environments folder if it doesn't exist
mkdir %USERPROFILE%\miniconda3\envs\deeplivecam

# Extract the package (Update the path to where you downloaded the file)
tar -xzf C:\path\to\deeplivecam.tar.gz -C %USERPROFILE%\miniconda3\envs\deeplivecam

# Activate the environment
conda activate deeplivecam

# Navigate to the app folder
cd Desktop\Deep-Live-Cam

# Run the GPU command
python run.py --execution-provider cuda --max-memory 5 --frame-processor face_swapper


## `--max-memory` Flag Explained

This flag prevents GPU crashes by limiting VRAM usage:

- `--max-memory 5` = Use max 5 GB (leaves 1 GB free for system)
- For 6 GB GPUs (RTX 4050/3060): Use `--max-memory 5`
- For 8 GB GPUs (RTX 4060/4070): Use `--max-memory 7`
- For 12 GB+ GPUs: Can omit or use `--max-memory 10`

**Why?** Without this limit, the program may try to use all available VRAM, causing Windows to crash the application.