Apollo-Windows 🌌

This is installation tutorial for using Apollo: Band-sequence Modeling for High-Quality Audio Restoration on Windows.

Tutorial for installing Apollo on Windows

⚠️ Note: The inference2.py file has been re-written to work on Windows systems.

Prerequisites

Before starting, make sure you have:

Installation

1. Clone the original Apollo repository

git clone https://github.com/JusperLee/Apollo
cd Apollo

2. Copy Windows-specific files

git clone https://github.com/starinspace/Apollo-Windows.git
copy Apollo-Windows\look2hear_win.yml .
copy Apollo-Windows\inference2.py .
copy Apollo-Windows\run.bat .
rmdir /s /q Apollo-Windows

3. Create a Conda environment

conda env create -f look2hear_win.yml -n look2hear_win

4. Activate a Conda environment

conda activate look2hear_win

5. Install PyTorch based on your GPU

Check your CUDA version:

nvidia-smi | findstr "CUDA Version"

Then install the correct PyTorch version:

🟢 CUDA 12.6

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

🟢 CUDA 12.8

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

🟢 CUDA 12.9

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu129

6. Install additional dependencies

Install Hugging Face Hub (optional):

pip install huggingface-hub
python -c "import huggingface_hub; print(huggingface_hub.__version__)"

Install any missing packages from the .yml file:

pip install omegaconf
conda install -c conda-forge ffmpeg
pip install "numpy<2"
pip install soundfile
pip install pydub
conda install -c conda-forge sox

7. Clone the model

git clone https://huggingface.co/JusperLee/Apollo

8. Create input/output folders

mkdir input
mkdir output

✅ You are now ready to run Apollo! Simply place your audio files in the input folder, run run.bat, and the processed files will appear in the output folder.

Apollo licensed under CC-BY-SA 4.0 International

starinspace/Apollo-Windows

README