Kokoro-ios Documentation

KokoroSwift is a TTS engine and does not bundle the large model files or voice styles. You are responsible for providing:

1. Model Files: The Kokoro TTS model files located at a valid local URL.
2. Voice Styles: Voice embeddings provided as MLXArray objects.

For guidance on how to bundle these files within an application, refer to the Kokoro Test App example project.

Add KokoroSwift to your project using Swift Package Manager (SPM). Add the repository URL to your dependencies and then link the KokoroSwift product to your target.

// In Package.swift
dependencies: [
    .package(url: "https://github.com/mlalma/kokoro-ios.git", from: "1.0.0")
]

// In your target definition
.target(
    name: "YourTarget",
    dependencies: [
        .product(name: "KokoroSwift", package: "kokoro-ios")
    ]
)

To use the TTS engine, initialize KokoroTTS with a local path to your model files and a G2P (Grapheme-to-Phoneme) option. You must provide a voice embedding as an MLXArray (which represents the voice style) to generate audio from text.

import KokoroSwift

// Initialize the TTS engine
let modelPath = URL(fileURLWithPath: "path/to/your/model")
let tts = KokoroTTS(modelPath: modelPath, g2p: .misaki)

// Generate speech
// Note: voiceEmbedding must be an `MLXArray` representing a voice style
let voiceEmbedding = ... 
let text = "Hello, this is a test of Kokoro TTS."
let audioBuffer = try tts.generateAudio(voice: voiceEmbedding, language: .enUS, text: text)

// audioBuffer now contains the synthesized speech

The KokoroTTS initializer accepts a g2p parameter to specify the Grapheme-to-Phoneme processor.

Supported options:

• .misaki: MisakiSwift (the default processor).
• .espeak: eSpeakNG (alternative processor, currently commented out in the current version).