Local Stem AI

Local-first audio separation studio for vocals, instrumentals, and music stems.

Private workflow. Clean dashboard. Local processing. No uploads, no account, no quota, no subscription.

Quick start · Features · Structure · Responsible use

At a glance

Area	Details
Workflow	Local audio separation from a browser dashboard
Focus	Vocals, instrumentals, and music stems
Privacy	Designed without hosted uploads or account requirements
Runtime	Generated files stay local and outside version control
Repository	Clean public baseline with screenshots, security policy, and automation files

Overview

Local Stem AI is a local dashboard for audio source separation workflows.

It helps you process your own audio files on your machine, follow separation jobs from a clear interface, and keep generated outputs organized in predictable local folders.

The repository is prepared as a clean public baseline: release screenshots are included, runtime folders are empty, generated files are excluded, and large model weights are intentionally kept out of version control.

Why Local Stem AI?

Many stem-separation workflows are either command-line utilities, hosted services, or subscription platforms.

Local Stem AI focuses on a different experience:

Principle	What it means
Local-first	Files are processed from your local workflow.
Privacy-conscious	Private recordings do not need to be uploaded to a hosted service.
Dashboard-driven	Jobs can be started and followed from a visual interface.
Repository-clean	Runtime data, logs, generated audio, and heavy assets stay out of the public repo.
Practical	The structure is easy to clone, inspect, run, and maintain.

What it does

Local Stem AI provides a dashboard-based workflow for separating audio sources.

Depending on the selected model, source file, and local environment, generated results may include:

• vocals
• instrumental versions
• separated music stems
• job-specific output folders
• locally generated result files

Built for

Local Stem AI is designed for users who want a private and practical local workflow:

• musicians preparing practice material
• producers working with authorized audio
• creators preparing local remix material
• vocal isolation workflows
• backing-track and karaoke preparation
• local audio experiments
• users who prefer not to upload private recordings to external platforms

Features

• local web dashboard
• audio separation workflow
• visual job tracking
• release screenshots included
• predictable runtime folder layout
• model configuration files included
• generated files excluded from version control
• large local assets excluded from version control
• GitHub Actions workflow included
• Dependabot configuration included
• security policy included

Workflow

Audio file
   ↓
Local dashboard
   ↓
Separation job
   ↓
Generated stems
   ↓
Local review and export

Project structure

.
├── dashboard/
│   ├── assets/
│   │   └── music_ai_treble.svg
│   ├── music_ai_dashboard.py
│   └── start_music_ai_dashboard.sh
├── docs/
│   └── screenshots/
│       ├── dashboard-home.png
│       └── job-page.png
├── input/
│   └── .gitkeep
├── output/
│   └── .gitkeep
├── logs/
│   └── .gitkeep
├── models/
│   └── audio-separator/
├── tmp/
│   └── .gitkeep
├── trash/
│   └── .gitkeep
├── README.md
├── requirements.txt
├── SECURITY.md
├── CHANGELOG.md
├── LICENSE
└── VERSION

Requirements

Recommended environment:

• Linux workstation
• Python 3
• dependencies from requirements.txt
• enough disk space for generated stems
• GPU acceleration when supported by the installed model stack

The repository can be cloned, reviewed, and audited without downloading large model weights or generating audio.

Quick start

Clone the repository:

git clone https://github.com/GhostInTheShell-444/local-stem-ai.git
cd local-stem-ai

Create a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Start the dashboard:

bash dashboard/start_music_ai_dashboard.sh

Open the local dashboard:

http://127.0.0.1:7865/

Runtime folders

The following folders are intentionally tracked only as empty placeholders:

input/
output/
logs/
tmp/
trash/

Generated files, logs, temporary files, audio outputs, and trash contents should remain local and should not be committed.

Model files

The repository includes model configuration files, not heavy model weights.

Large model files, checkpoints, generated outputs, and local runtime assets should stay outside version control. This keeps the project lightweight, easier to audit, and suitable for public distribution.

What is not included

Local Stem AI does not include:

• hosted audio processing
• cloud storage
• account management
• payment or credit systems
• bundled copyrighted audio
• generated audio outputs
• heavy model checkpoints
• private logs
• local runtime data

Responsible use

Local Stem AI is intended for audio you own or are authorized to process.

It does not provide downloading services, does not redistribute audio, and does not bypass copyright restrictions. Users are responsible for complying with applicable rights, licenses, and platform terms.

Security

Do not commit secrets, tokens, private logs, generated audio, local paths, or model weights.

See SECURITY.md for the project security policy.

License

This project is released under the license included in LICENSE.

Status

Local Stem AI is prepared as a clean public repository baseline for local audio separation workflows.