English | 日本語

swift-voice-input

Let people speak instead of type in an iOS or macOS app, with permission refusals, streaming partial text and start/stop state already handled.

Overview

• The engine is a seam, not a commitment — Apple's recogniser is the default, not the design. Anything conforming to SpeechRecognizer drops in, which is also what makes the state machine testable without a microphone.
• Text as it is spoken — partial results stream in and are published as a plain observable string, so a view binds to a property instead of managing a stream.
• Refused permissions handled — the one path most voice features get wrong. A denial is not retryable in-app, and the supplied button says so and routes the user to Settings.
• UI is optional — VoiceInput carries the recogniser and session; VoiceInputUI carries the SwiftUI controls. Depend on the first alone and SwiftUI never enters the build.

Usage

The host app must declare both usage descriptions in its Info.plist:

• NSMicrophoneUsageDescription — why the app records audio
• NSSpeechRecognitionUsageDescription — why that audio is transcribed

Missing either one terminates the app the moment that permission is requested, so the symptom is a crash on first tap rather than a denied permission.

import VoiceInput
import VoiceInputUI

struct DictationField: View {
    @State private var session = VoiceInputSession()
    @State private var text = ""

    var body: some View {
        HStack {
            TextField("Type here…", text: $text)
            VoiceInputButton(session: session)
        }
        .voiceInputOverlay(session: session) { transcript in
            text = transcript
        }
    }
}

Documentation

API reference and guides — including Getting Started, substituting a recognition engine, and what separates a partial result from a final one.

Requirements

iOS	macOS	Swift
17.0+	14.0+	6.2+

Installation

dependencies: [
    .package(url: "https://github.com/no-problem-dev/swift-voice-input.git", from: "3.0.0"),
]

Then take one or both products — VoiceInput for the session and recogniser, VoiceInputUI for the SwiftUI controls:

.product(name: "VoiceInput", package: "swift-voice-input"),
.product(name: "VoiceInputUI", package: "swift-voice-input"),

Contributing

See CONTRIBUTING.md.

License

MIT — see LICENSE.