Standalone Mode

To run Bitbox on a single computer, install both the backend processors and our Python library on that computer. See the installation guide for further details. A NVIDIA GPU must be present on that computer.

Setting up Backend Processors

Once the installation is complete, configure the Python library to connect with the backend processors. The simplest method is to utilize our Docker image and assign the image name to the runtime parameter.

# define a face processor and tell where the backend is located
processor = FP(runtime='bitbox:latest')

If you installed the backend processor natively, you can specify the installation directory.

# define a face processor and tell where the backend is located
processor = FP(runtime='/path/to/backend/bin')

To avoid setting the runtime every time you write a script, you can set the system environment variable in your local profile.

Mac / Linux

export BITBOX_DOCKER=bitbox:latest # if you want to use our Docker image
export BITBOX_3DI=/path/to/3DI/bin # if you installed 3DI natively
export BITBOX_3DI_LITE=/path/to/3DI-lite/bin # if you installed 3DI-lite natively

Windows

When the BITBOX_DOCKER environment variable is set, other environment variables are ignored. Meanwhile, setting the runtime parameter will disregard all environment variables, including BITBOX_DOCKER.

Pointing to Files

Define local input files and output directories using absolute or relative paths.

# define input file and output directory
input_file = 'data/elaine.mp4'
output_dir = 'output'

The output directory, if it doesn't already exist, will be automatically created. All files generated by the backend processor will be stored in this directory. Bitbox will monitor this directory for file caching; refer to our documentation for more details on file caching.