OSSSpeechKit

OSSSpeechKit is an iOS library for speech synthesis and speech recognition built on Apple's AVFoundation and Speech frameworks.

Requirements

• iOS 17 or later
• Swift 5 language mode or later
• Xcode with Swift 6 package support

Installation

In Xcode, choose File > Add Package Dependencies, enter:

https://github.com/AppDevGuy/OSSSpeechKit.git

Select version 1.0.0 or later and add the OSSSpeechKit library product to your app target.

You can also add it to Package.swift:

dependencies: [
    .package(
        url: "https://github.com/AppDevGuy/OSSSpeechKit.git",
        from: "1.0.0"
    )
]

Swift Package Manager is the only supported installation method. The historical CocoaPods release 0.3.3 remains available for existing builds but is no longer maintained; version 1.0 and later are not published to CocoaPods.

Speech synthesis

OSSSpeechEngine is main-actor isolated and instance based. Select a language from the catalog, confirm that a voice exists on the current device, then speak:

import AVFoundation
import OSSSpeechKit

let engine = OSSSpeechEngine()
let language = OSSLanguage.catalog.first {
    $0.id == "english-australia"
}!

guard !language.availableSynthesisVoices.isEmpty else {
    // Ask the user to install a compatible system voice.
    return
}

try await engine.speak(
    "Hello from OSSSpeechKit.",
    voice: OSSVoiceConfiguration(
        language: language,
        preferredQuality: .enhanced
    ),
    configuration: OSSUtteranceConfiguration(
        rate: AVSpeechUtteranceDefaultSpeechRate,
        pitchMultiplier: 1,
        volume: 1
    )
)

Use pauseSpeaking(), continueSpeaking(), and stopSpeaking() to control playback.

Speech recognition

Add both usage descriptions to the consuming app's Info.plist:

<key>NSMicrophoneUsageDescription</key>
<string>Explain why your app records audio.</string>
<key>NSSpeechRecognitionUsageDescription</key>
<string>Explain why your app transcribes speech.</string>

Then consume recognition events:

let engine = OSSSpeechEngine()

do {
    let events = try await engine.recognitionEvents(
        locale: Locale(identifier: "en-AU")
    )

    for try await event in events {
        switch event {
        case .partial(let text):
            print("Partial:", text)
        case .completed(let text):
            print("Final:", text)
        case .availabilityChanged(let available):
            print("Available:", available)
        case .cancelled:
            break
        }
    }
} catch {
    print("Recognition failed:", error)
}

Call cancelRecognition() to stop an active session. Set usesOnDeviceRecognition = true before starting if your product requires on-device recognition; the operation fails when the selected recognizer does not support it.

For a dictation-style session that stays open across verbal pauses, consume the continuous event stream:

let events = try await engine.continuousRecognitionEvents(
    locale: Locale(identifier: "en-AU")
)

for try await event in events {
    switch event {
    case .transcript(let transcript):
        print(transcript.formattedText)
        for segment in transcript.segments {
            print(segment.timestamp, segment.duration, segment.text)
        }
    case .paused:
        print("Capture paused")
    case .resumed:
        print("Capture resumed")
    case .completed, .cancelled:
        break
    case .availabilityChanged(let available):
        print("Available:", available)
    }
}

Use pauseRecognition() and resumeRecognition() to suspend and restart microphone capture while retaining the same event stream. Call finishRecognition() to end successfully, or cancelRecognition() to cancel. Segment timestamps use active recording time, so time spent paused is excluded. The API transcribes live microphone input; it does not save an audio file.

Languages, voices, and flags

OSSLanguage.catalog is display metadata, not a guarantee of runtime support. Availability varies by OS release, device, installed voice assets, region, network access, and Apple's services.

The catalog was reconciled against Apple's current VoiceOver language support and feature availability pages. Exact synthesis voices come from AVSpeechSynthesisVoice.speechVoices() and recognition locales come from SFSpeechRecognizer.supportedLocales(); those runtime results are authoritative.

Check capabilities when presenting an option:

for language in OSSLanguage.catalog {
    let canSpeak = !language.availableSynthesisVoices.isEmpty
    let canRecognize = language.supportsRecognition
    let flag = language.flagEmoji
    print(flag, language.name, canSpeak, canRecognize)
}

For UIKit code that needs an image, use:

let image = language.renderedFlagImage(pointSize: 24)

Speech synthesis is not translation. Supply text written for the selected language and locale.

Privacy

OSSSpeechKit requests microphone and speech-recognition permission only when your app asks it to. Your app is responsible for clear NSMicrophoneUsageDescription and NSSpeechRecognitionUsageDescription strings and for disclosing its own data practices.

Speech recognition may use Apple services unless on-device recognition is requested and supported. Review Apple's current Speech framework and App Store privacy requirements for your use case.

The included PrivacyInfo.xcprivacy declares that OSSSpeechKit itself does not track users, collect data, or access a required-reason API. It does not replace the consuming app's privacy manifest or App Privacy answers.

Migrating from pre-1.0

The singleton and subclass-based APIs remain temporarily available but are deprecated:

• OSSSpeech.shared → create an OSSSpeechEngine
• OSSVoice → OSSVoiceConfiguration
• OSSUtterance → OSSUtteranceConfiguration
• OSSVoiceEnum.flag → flagEmoji or renderedFlagImage(pointSize:scale:)
• OSSVoiceEnum catalogs → OSSLanguage.catalog

See MIGRATION.md for examples and behavior changes.

Example and tests

Open Example/OSSSpeechKit.xcodeproj to run the example app and its tests. The project links the package from the repository root.

License

OSSSpeechKit is available under the MIT license. See LICENSE.