Installation

1. Prerequisites

Clone the git repository to a local folder

git clone 'https://github.com/mcgrizzz/Koe.moe'
cd './Koemoe'

Goto the releases page and download the latest model. Place it into /koemoe/model folder that should be empty.

koemoe/

├── models/
│   └── [model].pt
├── condense.py
├── ...
└── utils.py

Download ffmpeg and install it

2. Continue with Conda or Pip

CondaPip

Download and install miniconda

In the main directory of Koemoe run the following commands

CUDACPU

conda env create -f 'environment_cuda.yml'
conda activate 'koemoe'

conda env create -f 'environment_cpu.yml'
conda activate 'koemoe'

Download and Install python 3.8+ In the main directory of Koemoe run the following commands

CUDACPU

pip install -r 'requirements_cuda.txt'

pip install -r 'requirements_cpu.txt'

Finally run one more command

pip install -r 'requirements_common.txt'

FAQ

Which do I install: CUDA or CPU?

graph LR
A(Do you have an Nvidia GPU?)-- YES -->B(CUDA);
A-- NO -->C(CPU);

Why use Conda?

Conda will create a self-contained environment for all of project requirements that is separate from the rest of your system. For example you can have python 3.8 installed on your system but in the conda environment it can be any other version. The downside to this approach is you must be in the conda environment (conda activate koemoe) in order to run the program