Skip to main content
This guide explains how the Voice to Text app implements speech recognition using Android’s built-in RecognizerIntent API with Jetpack Compose.

​
How it works

The app uses the ActivityResultContracts.StartActivityForResult() contract to launch the system’s speech recognition activity and receive the transcribed text.
1

Register the activity result launcher

Create a launcher that handles the speech recognition result using rememberLauncherForActivityResult.
val speechRecognizerLauncher = rememberLauncherForActivityResult(
    contract = ActivityResultContracts.StartActivityForResult(),
    onResult = { result ->
        val spokenText =
            result.data?.getStringArrayListExtra(RecognizerIntent.EXTRA_RESULTS)?.firstOrNull()
        if (spokenText != null) {
            prompt = spokenText  // Update prompt with recognized text
        } else {
            Toast.makeText(context, "Failed to recognize speech", Toast.LENGTH_SHORT).show()
        }
    }
)
The launcher extracts the first result from EXTRA_RESULTS and updates the UI state with the recognized text.
2

Create the recognition intent

Build an intent with RecognizerIntent.ACTION_RECOGNIZE_SPEECH and configure the language model and locale.
val intent = Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH)
intent.putExtra(
    RecognizerIntent.EXTRA_LANGUAGE_MODEL,
    RecognizerIntent.LANGUAGE_MODEL_FREE_FORM
)
intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, Locale.getDefault())
intent.putExtra(RecognizerIntent.EXTRA_PROMPT, "Speak now...")
LANGUAGE_MODEL_FREE_FORM is optimized for free-form speech rather than search queries.
3

Launch the recognizer

Call the launcher with the configured intent to start speech recognition.
speechRecognizerLauncher.launch(intent)
This opens the system’s speech recognition dialog where users can speak their input.

​
Intent configuration options

The RecognizerIntent API provides several configuration options: 
intent.putExtra(
    RecognizerIntent.EXTRA_LANGUAGE_MODEL,
    RecognizerIntent.LANGUAGE_MODEL_FREE_FORM
)
  • LANGUAGE_MODEL_FREE_FORM: For natural speech and dictation
  • LANGUAGE_MODEL_WEB_SEARCH: Optimized for short search queries
intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, Locale.getDefault())
Sets the recognition language. The app uses the device’s default locale, but you can specify any supported language (e.g., Locale.US, Locale.FRENCH).
intent.putExtra(RecognizerIntent.EXTRA_PROMPT, "Speak now...")
Displays a custom message in the speech recognition dialog.

​
Complete implementation

Here’s the full VoiceRecognitionScreen composable from the app: 
@Composable
fun VoiceRecognitionScreen(modifier: Modifier = Modifier) {
    val context = LocalContext.current
    var prompt by remember { mutableStateOf("") }

    // Launcher for speech recognition
    val speechRecognizerLauncher = rememberLauncherForActivityResult(
        contract = ActivityResultContracts.StartActivityForResult(),
        onResult = { result ->
            val spokenText =
                result.data?.getStringArrayListExtra(RecognizerIntent.EXTRA_RESULTS)?.firstOrNull()
            if (spokenText != null) {
                prompt = spokenText  // Update prompt with recognized text
            } else {
                Toast.makeText(context, "Failed to recognize speech", Toast.LENGTH_SHORT).show()
            }
        }
    )

    Box(
        modifier = modifier.fillMaxSize(),
        contentAlignment = Alignment.Center
    ) {
        Row(
            verticalAlignment = Alignment.CenterVertically,
            modifier = Modifier
                .fillMaxWidth()
                .padding(horizontal = 16.dp)
        ) {
            BasicTextField(
                value = prompt,
                onValueChange = { prompt = it },
                modifier = Modifier
                    .weight(1f)
                    .padding(8.dp)
                    .border(1.dp, MaterialTheme.colorScheme.primary)
                    .padding(8.dp),
                singleLine = true,
                decorationBox = { innerTextField ->
                    if (prompt.isEmpty()) {
                        Text("Type or speak your message...", color = Color.Gray)
                    }
                    innerTextField()
                }
            )

            Button(
                onClick = {
                    if (ContextCompat.checkSelfPermission(
                            context,
                            Manifest.permission.RECORD_AUDIO
                        ) == PackageManager.PERMISSION_GRANTED
                    ) {
                        val intent = Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH)
                        intent.putExtra(
                            RecognizerIntent.EXTRA_LANGUAGE_MODEL,
                            RecognizerIntent.LANGUAGE_MODEL_FREE_FORM
                        )
                        intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, Locale.getDefault())
                        intent.putExtra(RecognizerIntent.EXTRA_PROMPT, "Speak now...")
                        speechRecognizerLauncher.launch(intent)
                    } else {
                        ActivityCompat.requestPermissions(
                            context as Activity,
                            arrayOf(Manifest.permission.RECORD_AUDIO),
                            100
                        )
                    }
                },
                modifier = Modifier.padding(start = 8.dp)
            ) {
                Text("Speak")
            }
        }
    }
}

​
Key implementation details

The speech recognizer launcher must be created at the composable level using rememberLauncherForActivityResult. You cannot create it inside the button’s onClick handler.
Always check for RECORD_AUDIO permission before launching the speech recognizer. See the permissions guide for details.

​
Error handling

The app handles recognition failures by checking if spokenText is null: 
if (spokenText != null) {
    prompt = spokenText
} else {
    Toast.makeText(context, "Failed to recognize speech", Toast.LENGTH_SHORT).show()
}
Common failure scenarios include:
  • User cancels the recognition dialog
  • No speech detected
  • Network issues (for cloud-based recognizers)
  • Recognizer not available on the device

Build docs developers (and LLMs) love

Get started for freeTalk to us