Credit to Original Repositories

• Original repo: https://github.com/Audio-AGI/AudioSep
• camenduru's colab with the gradio UI: https://github.com/camenduru/AudioSep-colab

This repo is forked from camenduru's colab code to install locally.

Installation

1. Create new conda environment

conda create -n audiosep python=3.10
conda activate audiosep

2. Clone the repository

git clone https://github.com/natlamir/AudioSep.git
cd AudioSep

3. Install dependencies

pip install -r requirements.txt

4. Download required files

• model weights: place in the checkpoint folder
• audioset_textmap.npy: place in the models/CLAP/training folder
• bpe_simple_vocab_16e6.txt.gz: place in the models/CLAP/open_clip folder

Usage

Run the gradio app with:

python app.py