jiyimeta/swift-spleeter

Features

• Uses Deezer's Spleeter model compiled as a Core ML model
• Supports streaming audio processing for long files
• Exposes separation results as monaural audio
• Works with stereo input (or mono, automatically handled)
• Async/Await and progress-reporting support (macOS 15+, iOS 18+)

Installation

To use the Spleeter library in a SwiftPM project, add the following line to the dependencies in your Package.swift file:

.package(url: "https://github.com/jiyimeta/swift-spleeter", from: "0.1.0"),

Include "Spleeter" as a dependency for your executable target:

.target(name: "<target>", dependencies: [
    .product(name: "Spleeter", package: "swift-spleeter"),
]),

Finally, add import Spleeter to your source code.

Usage

1. Separate from file to file

import SwiftSpleeter

let modelURL = URL(fileURLWithPath: "path/to/Spleeter2.mlmodelc")
let inputURL = URL(fileURLWithPath: "path/to/mixture.wav")
let outputURLs = Stems2(
    vocals: FileManager.default.temporaryDirectory.appendingPathComponent("vocals.wav"),
    instruments: FileManager.default.temporaryDirectory.appendingPathComponent("instruments.wav")
)

let separator = try AudioSeparator2(modelURL: modelURL)

for try await progress in separator.separate(from: inputURL, to: outputURLs) {
    print("Progress: \(progress.current)/\(progress.total)")
}

2. Separate in-memory and handle results

import SwiftSpleeter

let modelURL = URL(fileURLWithPath: "path/to/Spleeter2.mlmodelc")
let inputURL = URL(fileURLWithPath: "path/to/mixture.wav")

let separator = try AudioSeparator2(modelURL: modelURL)
let file = try AudioFile(forReading: inputURL)
let samples = try file.readStereoSamples()

for try await (result, progress) in separator.separate(samples) {
    if let result = result {
        // result.vocals: [Float], result.instruments: [Float]
        print("Received chunk: vocals.count = \(result.vocals.count)")
    }
    print("Progress: \(progress.current)/\(progress.total)")
}

Requirements

• A compiled .mlmodelc version of the 2-, 4- or 5-stem Spleeter model
• Designed to work on macOS 15 / iOS 18 / tvOS 18 / watchOS 11 / visionOS 2 or later
• Library can be imported on earlier OS versions, but usage requires runtime availability checks via if #available(...)

Models

• A compiled model (Spleeter2Model.mlmodelc) is included in Demo App and can be used directly.
• To customize the model, use the provided spleeter-pytorch submodule under Tools/spleeter-pytorch/ to export and compile .mlmodelc files.

Demo App

To try the library in action:

cd Example
xcodegen
open SpleeterExample.xcodeproj

This will generate and open a demo Xcode project that demonstrates the separation functionality.

License

This library is released under the MIT license. See LICENSE for details.

Acknowledgements

This project uses the 2-stem Spleeter model by Deezer Research.

Package Metadata

Repository: jiyimeta/swift-spleeter

Default branch: main

README: README.md