metadata

title: Tune Splitter III
emoji: 🐨
colorFrom: green
colorTo: green
sdk: streamlit
sdk_version: 1.44.1
app_file: app.py
pinned: false
short_description: An agent for splitting song stems

Music Stem Splitter

This Streamlit application separates music tracks into individual stems (vocals, drums, bass, and other instruments) using the Demucs AI model.

Features

Upload MP3, WAV, FLAC, or OGG audio files
Separate audio into four stems: vocals, drums, bass, and other instruments
Listen to separated stems directly in the browser
View spectrograms for each stem
Download individual stems for further use

Demo

Upload an audio file using the file uploader
Click "Split into Stems" to process the audio
Navigate through the tabs to listen to each stem
Download stems individually with the provided links

Technical Details

This application uses the Demucs v4 model (HT-Demucs) developed by Facebook AI Research to perform music source separation. The model leverages deep neural networks to identify and isolate different components of a music recording.

Limitations

Due to Hugging Face Spaces resource constraints:

Maximum file size: 100MB
Maximum audio duration: 5 minutes
Processing may take several minutes depending on server load

Local Installation

To run this application locally:

git clone https://huggingface.co/spaces/username/music-stem-splitter
cd music-stem-splitter
pip install -r requirements.txt
streamlit run app.py

Credits

Demucs by Facebook AI Research
Built with Streamlit
Hosted on Hugging Face Spaces