@iternio/react-native-auto-play 0.5.10 → 0.5.12

package/android/src/main/java/com/margelo/nitro/swe/iternio/reactnativeautoplay/HybridVoice.kt

@@ -70,7 +70,8 @@ class HybridVoice : HybridVoiceSpec() {
         onChunk: ((chunk: VoiceInputChunk) -> Unit)?,
         language: String?,
         startSoundUri: String?,
-        endSoundUri: String?
+        endSoundUri: String?,
+        encoding: VoiceAudioEncoding?
     ): Promise<VoiceInputResult> {
         return Promise.async {
             if (Build.VERSION.SDK_INT < Build.VERSION_CODES.O) {
@@ -89,6 +90,7 @@ class HybridVoice : HybridVoiceSpec() {
                     language = language,
                     startSoundUri = startSoundUri,
                     endSoundUri = endSoundUri,
+                    encoding = encoding ?: VoiceAudioEncoding.LINEAR16,
                 )
             } finally {
                 voiceInputManager = null

package/android/src/main/java/com/margelo/nitro/swe/iternio/reactnativeautoplay/VoiceInputManager.kt

@@ -27,6 +27,7 @@ import androidx.core.net.toUri
 import com.facebook.react.bridge.UiThreadUtil
 import com.margelo.nitro.NitroModules
 import com.margelo.nitro.core.ArrayBuffer
+import com.margelo.nitro.swe.iternio.reactnativeautoplay.utils.G711
 import com.margelo.nitro.swe.iternio.reactnativeautoplay.utils.ThreadUtil
 import kotlinx.coroutines.CoroutineScope
 import kotlinx.coroutines.Dispatchers
@@ -81,6 +82,7 @@ class VoiceInputManager(
         language: String? = null,
         startSoundUri: String? = null,
         endSoundUri: String? = null,
+        encoding: VoiceAudioEncoding = VoiceAudioEncoding.LINEAR16,
     ): VoiceInputResult {
         cancelledByUser = false
         if (!requestAudioFocus()) {
@@ -93,19 +95,19 @@ class VoiceInputManager(
                 if (SpeechRecognizer.isRecognitionAvailable(context)) {
                     if (carContext != null) {
                         if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.TIRAMISU) {
-                            startSTTFromCarAudio(silenceThresholdMs, maxDurationMs, onChunk, language)
+                            startSTTFromCarAudio(silenceThresholdMs, maxDurationMs, encoding, onChunk, language)
                         } else {
                             // Car connected but API < 33: EXTRA_AUDIO_SOURCE unavailable, fall back to PCM
-                            startPCM(silenceThresholdMs, maxDurationMs, onChunk)
+                            startPCM(silenceThresholdMs, maxDurationMs, encoding, onChunk)
                         }
                     } else {
                         ThreadUtil.postOnUiAndAwait { startSTT(context, onChunk, language) }.getOrThrow()
                     }
                 } else {
-                    startPCM(silenceThresholdMs, maxDurationMs, onChunk)
+                    startPCM(silenceThresholdMs, maxDurationMs, encoding, onChunk)
                 }
             } else {
-                startPCM(silenceThresholdMs, maxDurationMs, onChunk)
+                startPCM(silenceThresholdMs, maxDurationMs, encoding, onChunk)
             }
             startSoundJob?.join()
             if (cancelledByUser) throw VoiceInputCancelledException()
@@ -182,6 +184,7 @@ class VoiceInputManager(
     private suspend fun startSTTFromCarAudio(
         silenceThresholdMs: Long,
         maxDurationMs: Long,
+        encoding: VoiceAudioEncoding,
         onChunk: ((chunk: VoiceInputChunk) -> Unit)?,
         language: String?
     ): VoiceInputResult {
@@ -225,7 +228,9 @@ class VoiceInputManager(
         return try {
             sttDeferred.await()
         } catch (_: Exception) {
-            val directBuffer = ByteBuffer.allocateDirect(pcmBytes.size).put(pcmBytes).rewind() as ByteBuffer
+            val encodedBytes = encodeBytes(pcmBytes, encoding)
+            val directBuffer =
+                ByteBuffer.allocateDirect(encodedBytes.size).put(encodedBytes).rewind() as ByteBuffer
             VoiceInputResult(transcription = null, audio = ArrayBuffer.wrap(directBuffer))
         }
     }
@@ -315,14 +320,36 @@ class VoiceInputManager(
     private suspend fun startPCM(
         silenceThresholdMs: Long,
         maxDurationMs: Long,
+        encoding: VoiceAudioEncoding,
         onChunk: ((chunk: VoiceInputChunk) -> Unit)?,
     ): VoiceInputResult {
-        val pcmBytes = recordPCM(silenceThresholdMs, maxDurationMs, onChunk)
+        // Only the plain-PCM result/onChunk are re-encoded — STT paths (startSTT,
+        // startSTTFromCarAudio) always stay LINEAR16 since they feed the recognizer directly.
+        val wrappedOnChunk: ((VoiceInputChunk) -> Unit)? = onChunk?.let { cb ->
+            { chunk -> cb(encodeChunk(chunk, encoding)) }
+        }
+        val pcmBytes = recordPCM(silenceThresholdMs, maxDurationMs, wrappedOnChunk)
+        val encodedBytes = encodeBytes(pcmBytes, encoding)
         val directBuffer =
-            ByteBuffer.allocateDirect(pcmBytes.size).put(pcmBytes).rewind() as ByteBuffer
+            ByteBuffer.allocateDirect(encodedBytes.size).put(encodedBytes).rewind() as ByteBuffer
         return VoiceInputResult(transcription = null, audio = ArrayBuffer.wrap(directBuffer))
     }
 
+    private fun encodeBytes(pcm16le: ByteArray, encoding: VoiceAudioEncoding): ByteArray =
+        when (encoding) {
+            VoiceAudioEncoding.LINEAR16 -> pcm16le
+            VoiceAudioEncoding.MULAW -> G711.encodeUlaw(pcm16le)
+            VoiceAudioEncoding.ALAW -> G711.encodeAlaw(pcm16le)
+        }
+
+    private fun encodeChunk(chunk: VoiceInputChunk, encoding: VoiceAudioEncoding): VoiceInputChunk {
+        val audio = chunk.audio
+        if (audio == null || encoding == VoiceAudioEncoding.LINEAR16) return chunk
+        val encoded = encodeBytes(audio.toByteArray(), encoding)
+        val direct = ByteBuffer.allocateDirect(encoded.size).put(encoded).rewind() as ByteBuffer
+        return VoiceInputChunk(partial = chunk.partial, audio = ArrayBuffer.wrap(direct))
+    }
+
     @SuppressLint("MissingPermission")
     @RequiresApi(Build.VERSION_CODES.O)
     private suspend fun recordPCM(
@@ -365,6 +392,16 @@ class VoiceInputManager(
         }
 
         val outputStream = ByteArrayOutputStream()
+        val pendingChunk = onChunk?.let { ByteArrayOutputStream() }
+
+        fun flushPendingChunk() {
+            val pending = pendingChunk ?: return
+            if (pending.size() == 0) return
+            val bytes = pending.toByteArray()
+            pending.reset()
+            val direct = ByteBuffer.allocateDirect(bytes.size).put(bytes).rewind() as ByteBuffer
+            onChunk?.invoke(VoiceInputChunk(partial = null, audio = ArrayBuffer.wrap(direct)))
+        }
 
         recordingJob = scope.launch {
             val buffer = ByteArray(bufferSize)
@@ -380,19 +417,22 @@ class VoiceInputManager(
                     ) ?: -1
 
                     if (read < 0) {
-                        // Whenever the user dismisses the microphone on the car screen, the next call to read will return -1
-                        cancelledByUser = carAudioRecord != null && read == -1
+                        // Whenever the user dismisses the microphone on the car screen, the next call to read will return -1.
+                        // But calling stop() ourselves also races CarAudioRecord's internal stream close against an
+                        // in-flight read(), which can likewise surface as -1 — isRecording is already false by then
+                        // (set before stopRecording() is called), so it distinguishes the two cases.
+                        cancelledByUser = carAudioRecord != null && read == -1 && isRecording
                         break
                     }
 
                     if (read > 0) {
                         outputStream.write(buffer, 0, read)
 
-                        onChunk?.let { cb ->
-                            val chunk = ByteArray(read) { buffer[it] }
-                            val direct =
-                                ByteBuffer.allocateDirect(read).put(chunk).rewind() as ByteBuffer
-                            cb(VoiceInputChunk(partial = null, audio = ArrayBuffer.wrap(direct)))
+                        pendingChunk?.let { pending ->
+                            pending.write(buffer, 0, read)
+                            if (pending.size() >= CHUNK_EMIT_BYTES) {
+                                flushPendingChunk()
+                            }
                         }
 
                         val now = System.currentTimeMillis()
@@ -430,6 +470,7 @@ class VoiceInputManager(
                     }
                 }
             } finally {
+                flushPendingChunk()
                 releaseResources()
                 val captured = pcmContinuation
                 pcmContinuation = null
@@ -439,6 +480,10 @@ class VoiceInputManager(
     }
 
     fun stop() {
+        // Mark as no longer recording before triggering any teardown side effects, so anything
+        // racing against a concurrent read() sees this is an app-initiated stop.
+        isRecording = false
+
         // STT path: stopListening() triggers onResults/onError which resolves the continuation
         activeSpeechRecognizer?.let { recognizer ->
             UiThreadUtil.runOnUiThread {
@@ -446,7 +491,6 @@ class VoiceInputManager(
             }
         }
         // PCM path and car-audio STT pump
-        isRecording = false
         carAudioRecord?.stopRecording()
         audioRecord?.stop()
     }
@@ -552,6 +596,7 @@ class VoiceInputManager(
         private const val WARMUP_MS = 500L
         private const val SAMPLE_RATE = 16_000
         private const val PHONE_BUFFER_SIZE = 3_200 // ~100ms at 16kHz/16-bit/mono
+        private const val CHUNK_EMIT_BYTES = 3_200 // ~100ms at 16kHz/16-bit/mono, batches onChunk callbacks
 
         fun hasVoiceInputPermission(): Boolean {
             val context = NitroModules.applicationContext ?: return false

package/android/src/main/java/com/margelo/nitro/swe/iternio/reactnativeautoplay/utils/G711.kt

@@ -0,0 +1,77 @@
+package com.margelo.nitro.swe.iternio.reactnativeautoplay.utils
+
+/** Encodes 16-bit signed little-endian PCM to G.711 µ-law/A-law (ITU-T reference algorithm). */
+object G711 {
+    private const val ULAW_BIAS = 0x84
+    private const val ULAW_CLIP = 8159
+
+    private val ULAW_EXP_LUT = intArrayOf(
+        0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
+        5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+        6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+        6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+    )
+
+    private val ALAW_SEG_END = intArrayOf(0x1F, 0x3F, 0x7F, 0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF)
+
+    private fun search(value: Int, table: IntArray): Int {
+        for (i in table.indices) {
+            if (value <= table[i]) return i
+        }
+        return table.size
+    }
+
+    private fun linearToUlaw(sampleIn: Int): Byte {
+        var sample = sampleIn
+        val sign = (sample shr 8) and 0x80
+        if (sign != 0) sample = -sample
+        if (sample > ULAW_CLIP) sample = ULAW_CLIP
+        sample += ULAW_BIAS
+        val exponent = ULAW_EXP_LUT[(sample shr 7) and 0xFF]
+        val mantissa = (sample shr (exponent + 3)) and 0x0F
+        var ulawByte = (sign or (exponent shl 4) or mantissa).inv()
+        if (ulawByte == 0) ulawByte = 0x02
+        return ulawByte.toByte()
+    }
+
+    private fun linearToAlaw(sampleIn: Int): Byte {
+        var sample = sampleIn shr 3
+        val mask: Int
+        if (sample >= 0) {
+            mask = 0xD5
+        } else {
+            mask = 0x55
+            sample = -sample - 1
+        }
+
+        val seg = search(sample, ALAW_SEG_END)
+        val alawByte = if (seg >= 8) {
+            0x7F xor mask
+        } else {
+            var aval = seg shl 4
+            aval = if (seg < 2) aval or ((sample shr 1) and 0x0F) else aval or ((sample shr seg) and 0x0F)
+            aval xor mask
+        }
+        return alawByte.toByte()
+    }
+
+    /** Encodes 16-bit signed little-endian PCM bytes to 8-bit G.711. Drops a trailing odd byte, if any. */
+    fun encode(pcm16le: ByteArray, encoding: (Int) -> Byte): ByteArray {
+        val sampleCount = pcm16le.size / 2
+        val out = ByteArray(sampleCount)
+        for (i in 0 until sampleCount) {
+            val lo = pcm16le[i * 2].toInt() and 0xFF
+            val hi = pcm16le[i * 2 + 1].toInt()
+            val sample = (hi shl 8) or lo
+            out[i] = encoding(sample)
+        }
+        return out
+    }
+
+    fun encodeUlaw(pcm16le: ByteArray): ByteArray = encode(pcm16le, ::linearToUlaw)
+    fun encodeAlaw(pcm16le: ByteArray): ByteArray = encode(pcm16le, ::linearToAlaw)
+}

package/ios/hybrid/HybridVoice.swift

@@ -38,7 +38,8 @@ class HybridVoice: HybridVoiceSpec {
         onChunk: ((_ chunk: VoiceInputChunk) -> Void)?,
         language: String?,
         startSoundUri: String?,
-        endSoundUri: String?
+        endSoundUri: String?,
+        encoding: VoiceAudioEncoding?
     ) throws -> Promise<VoiceInputResult> {
         return Promise.async {
             let interfaceController = try? await RootModule.withInterfaceController { $0 }
@@ -59,7 +60,8 @@ class HybridVoice: HybridVoiceSpec {
                 onChunk: onChunk,
                 language: language,
                 startSoundUri: startSoundUri,
-                endSoundUri: endSoundUri
+                endSoundUri: endSoundUri,
+                encoding: encoding ?? .linear16
             )
         }
     }

package/ios/utils/G711.swift

@@ -0,0 +1,83 @@
+import Foundation
+
+/// Encodes 16-bit signed little-endian PCM to G.711 µ-law/A-law (ITU-T reference algorithm).
+enum G711 {
+    private static let ulawBias = 0x84
+    private static let ulawClip = 8159
+
+    private static let ulawExpLut: [Int] = [
+        0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
+        5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+        6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+        6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+        7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+    ]
+
+    private static let alawSegEnd = [0x1F, 0x3F, 0x7F, 0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF]
+
+    private static func search(_ value: Int, _ table: [Int]) -> Int {
+        for (i, boundary) in table.enumerated() where value <= boundary {
+            return i
+        }
+        return table.count
+    }
+
+    private static func linearToUlaw(_ sampleIn: Int) -> UInt8 {
+        var sample = sampleIn
+        let sign = (sample >> 8) & 0x80
+        if sign != 0 { sample = -sample }
+        if sample > ulawClip { sample = ulawClip }
+        sample += ulawBias
+        let exponent = ulawExpLut[(sample >> 7) & 0xFF]
+        let mantissa = (sample >> (exponent + 3)) & 0x0F
+        var ulawByte = ~(sign | (exponent << 4) | mantissa)
+        ulawByte &= 0xFF
+        if ulawByte == 0 { ulawByte = 0x02 }
+        return UInt8(ulawByte)
+    }
+
+    private static func linearToAlaw(_ sampleIn: Int) -> UInt8 {
+        var sample = sampleIn >> 3
+        let mask: Int
+        if sample >= 0 {
+            mask = 0xD5
+        }
+        else {
+            mask = 0x55
+            sample = -sample - 1
+        }
+
+        let seg = search(sample, alawSegEnd)
+        let alawByte: Int
+        if seg >= 8 {
+            alawByte = 0x7F ^ mask
+        }
+        else {
+            var aval = seg << 4
+            aval |= seg < 2 ? (sample >> 1) & 0x0F : (sample >> seg) & 0x0F
+            alawByte = aval ^ mask
+        }
+        return UInt8(alawByte & 0xFF)
+    }
+
+    /// Encodes 16-bit signed little-endian PCM data to 8-bit G.711. Drops a trailing odd byte, if any.
+    private static func encode(_ pcm16le: Data, _ encoding: (Int) -> UInt8) -> Data {
+        let sampleCount = pcm16le.count / 2
+        var out = [UInt8](repeating: 0, count: sampleCount)
+        pcm16le.withUnsafeBytes { (raw: UnsafeRawBufferPointer) in
+            for i in 0..<sampleCount {
+                let lo = Int(raw[i * 2])
+                let hi = Int(Int8(bitPattern: raw[i * 2 + 1]))
+                let sample = (hi << 8) | lo
+                out[i] = encoding(sample)
+            }
+        }
+        return Data(out)
+    }
+
+    static func encodeUlaw(_ pcm16le: Data) -> Data { encode(pcm16le, linearToUlaw) }
+    static func encodeAlaw(_ pcm16le: Data) -> Data { encode(pcm16le, linearToAlaw) }
+}

package/ios/utils/VoiceInputManager.swift

@@ -69,6 +69,9 @@ class VoiceInputManager {
     private var silenceStart: Date?
     private var firstBufferContinuation: CheckedContinuation<Void, Never>?
 
+    // PCM result/onChunk audio encoding — STT transcription itself is unaffected
+    private var encoding: VoiceAudioEncoding = .linear16
+
     private static let sampleRate: Double = 16_000
     private static let tapBufferSize: AVAudioFrameCount = 4_096
     private static let silenceAmplitudeThreshold = 500
@@ -94,8 +97,10 @@ class VoiceInputManager {
         onChunk: ((_ chunk: VoiceInputChunk) -> Void)?,
         language: String?,
         startSoundUri: String?,
-        endSoundUri: String?
+        endSoundUri: String?,
+        encoding: VoiceAudioEncoding
     ) async throws -> VoiceInputResult {
+        self.encoding = encoding
         stopLock.withLock {
             cancelledByUser = false
         }
@@ -355,9 +360,8 @@ class VoiceInputManager {
 
             // PCM chunk callback
             if activeRecognitionRequest == nil, let onChunk {
-                if let chunkBuffer = try? ArrayBuffer.copy(
-                    data: newSamples.withUnsafeBufferPointer { Data(buffer: $0) }
-                ) {
+                let pcmData = newSamples.withUnsafeBufferPointer { Data(buffer: $0) }
+                if let chunkBuffer = try? ArrayBuffer.copy(data: self.encodeAudio(pcmData)) {
                     onChunk(VoiceInputChunk(partial: nil, audio: chunkBuffer))
                 }
             }
@@ -424,10 +428,18 @@ class VoiceInputManager {
 
     private func makePCMResult(from samples: [Int16]) -> VoiceInputResult {
         let data = samples.withUnsafeBufferPointer { Data(buffer: $0) }
-        let buffer = try? ArrayBuffer.copy(data: data)
+        let buffer = try? ArrayBuffer.copy(data: encodeAudio(data))
         return VoiceInputResult(transcription: nil, audio: buffer)
     }
 
+    private func encodeAudio(_ pcm16le: Data) -> Data {
+        switch encoding {
+        case .linear16: return pcm16le
+        case .mulaw: return G711.encodeUlaw(pcm16le)
+        case .alaw: return G711.encodeAlaw(pcm16le)
+        }
+    }
+
     // CPVoiceControlState enforces a maximum image size of 150x150 points.
     private static let voiceImageMaxSize = CGSize(width: 150, height: 150)
 

package/lib/hybrid/HybridVoice.d.ts

@@ -39,6 +39,7 @@ export declare const HybridVoice: {
      * @param preferSpeechToText  request STT transcription instead of raw PCM
      * @param onChunk             optional streaming callback
      * @param language            specify the language for the SpeechRecognizer, falls back to system language if not set
+     * @param encoding            PCM encoding for onChunk audio and the final result (default LINEAR16)
      */
     startVoiceInput: StartVoiceInput;
     /**

package/lib/hybrid/HybridVoice.js

@@ -3,11 +3,11 @@ import { NitroModules } from 'react-native-nitro-modules';
 import { NitroImageUtil } from '../utils/NitroImage';
 const _native = NitroModules.createHybridObject('Voice');
 const startVoiceInput = async (options) => {
-    const { onChunk, silenceThresholdMs, maxDurationMs, listeningText, listeningImage, preferSpeechToText, language, startSound, endSound, } = options ?? {};
+    const { onChunk, silenceThresholdMs, maxDurationMs, listeningText, listeningImage, preferSpeechToText, language, startSound, endSound, encoding, } = options ?? {};
     const listeningImageRepeats = listeningImage?.type === 'asset' ? listeningImage.repeats : undefined;
     const startSoundUri = startSound != null ? Image.resolveAssetSource(startSound).uri : undefined;
     const endSoundUri = endSound != null ? Image.resolveAssetSource(endSound).uri : undefined;
-    return await _native.startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, NitroImageUtil.convert(listeningImage), listeningImageRepeats, preferSpeechToText, onChunk, language, startSoundUri, endSoundUri);
+    return await _native.startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, NitroImageUtil.convert(listeningImage), listeningImageRepeats, preferSpeechToText, onChunk, language, startSoundUri, endSoundUri, encoding);
 };
 export const HybridVoice = {
     /**
@@ -45,6 +45,7 @@ export const HybridVoice = {
      * @param preferSpeechToText  request STT transcription instead of raw PCM
      * @param onChunk             optional streaming callback
      * @param language            specify the language for the SpeechRecognizer, falls back to system language if not set
+     * @param encoding            PCM encoding for onChunk audio and the final result (default LINEAR16)
      */
     startVoiceInput,
     /**

package/lib/specs/Voice.nitro.d.ts

@@ -1,5 +1,5 @@
 import type { HybridObject } from 'react-native-nitro-modules';
-import type { VoiceInputChunk, VoiceInputResult } from '../types/Voice';
+import type { VoiceAudioEncoding, VoiceInputChunk, VoiceInputResult } from '../types/Voice';
 import type { NitroImage } from '../utils/NitroImage';
 export interface Voice extends HybridObject<{
     android: 'kotlin';
@@ -7,6 +7,6 @@ export interface Voice extends HybridObject<{
 }> {
     hasVoiceInputPermission(): boolean;
     requestVoiceInputPermission(): Promise<boolean>;
-    startVoiceInput(silenceThresholdMs?: number, maxDurationMs?: number, listeningText?: string, listeningImage?: NitroImage, listeningImageRepeats?: boolean, preferSpeechToText?: boolean, onChunk?: (chunk: VoiceInputChunk) => void, language?: string, startSoundUri?: string, endSoundUri?: string): Promise<VoiceInputResult>;
+    startVoiceInput(silenceThresholdMs?: number, maxDurationMs?: number, listeningText?: string, listeningImage?: NitroImage, listeningImageRepeats?: boolean, preferSpeechToText?: boolean, onChunk?: (chunk: VoiceInputChunk) => void, language?: string, startSoundUri?: string, endSoundUri?: string, encoding?: VoiceAudioEncoding): Promise<VoiceInputResult>;
     stopVoiceInput(): void;
 }

package/lib/types/Voice.d.ts

@@ -1,4 +1,8 @@
 import type { VoiceInputImage } from './Image';
+/** PCM sample encoding for the `audio` payload of chunks and the final result.
+ * `LINEAR16` is raw 16-bit signed PCM
+ * `MULAW`/`ALAW` are G.711 8-bit companded encodings, halving payload size */
+export type VoiceAudioEncoding = 'LINEAR16' | 'MULAW' | 'ALAW';
 export interface VoiceInputChunk {
     partial?: string;
     audio?: ArrayBuffer;
@@ -18,6 +22,8 @@ export interface VoiceInputOptions {
     preferSpeechToText?: boolean;
     onChunk?: (chunk: VoiceInputChunk) => void;
     language?: string;
+    /** PCM encoding for onChunk audio and the final result. Defaults to LINEAR16. */
+    encoding?: VoiceAudioEncoding;
     /** Sound played just before recording starts. Pass a Metro asset: `require('./beep_start.wav')`. */
     startSound?: number;
     /** Sound played just after recording stops. Pass a Metro asset: `require('./beep_end.wav')`. */

package/nitrogen/generated/android/c++/JHybridVoiceSpec.cpp

@@ -19,6 +19,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct RemoteImage
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct NitroColor; }
 // Forward declaration of `VoiceInputChunk` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputChunk; }
+// Forward declaration of `VoiceAudioEncoding` to properly resolve imports.
+namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VoiceAudioEncoding; }
 
 #include <NitroModules/Promise.hpp>
 #include <NitroModules/JPromise.hpp>
@@ -43,6 +45,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputC
 #include "JFunc_void_VoiceInputChunk.hpp"
 #include <NitroModules/JNICallable.hpp>
 #include "JVoiceInputChunk.hpp"
+#include "VoiceAudioEncoding.hpp"
+#include "JVoiceAudioEncoding.hpp"
 
 namespace margelo::nitro::swe::iternio::reactnativeautoplay {
 
@@ -98,9 +102,9 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay {
       return __promise;
     }();
   }
-  std::shared_ptr<Promise<VoiceInputResult>> JHybridVoiceSpec::startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri) {
-    static const auto method = _javaPart->javaClassStatic()->getMethod<jni::local_ref<JPromise::javaobject>(jni::alias_ref<jni::JDouble> /* silenceThresholdMs */, jni::alias_ref<jni::JDouble> /* maxDurationMs */, jni::alias_ref<jni::JString> /* listeningText */, jni::alias_ref<JVariant_GlyphImage_AssetImage_RemoteImage> /* listeningImage */, jni::alias_ref<jni::JBoolean> /* listeningImageRepeats */, jni::alias_ref<jni::JBoolean> /* preferSpeechToText */, jni::alias_ref<JFunc_void_VoiceInputChunk::javaobject> /* onChunk */, jni::alias_ref<jni::JString> /* language */, jni::alias_ref<jni::JString> /* startSoundUri */, jni::alias_ref<jni::JString> /* endSoundUri */)>("startVoiceInput_cxx");
-    auto __result = method(_javaPart, silenceThresholdMs.has_value() ? jni::JDouble::valueOf(silenceThresholdMs.value()) : nullptr, maxDurationMs.has_value() ? jni::JDouble::valueOf(maxDurationMs.value()) : nullptr, listeningText.has_value() ? jni::make_jstring(listeningText.value()) : nullptr, listeningImage.has_value() ? JVariant_GlyphImage_AssetImage_RemoteImage::fromCpp(listeningImage.value()) : nullptr, listeningImageRepeats.has_value() ? jni::JBoolean::valueOf(listeningImageRepeats.value()) : nullptr, preferSpeechToText.has_value() ? jni::JBoolean::valueOf(preferSpeechToText.value()) : nullptr, onChunk.has_value() ? JFunc_void_VoiceInputChunk_cxx::fromCpp(onChunk.value()) : nullptr, language.has_value() ? jni::make_jstring(language.value()) : nullptr, startSoundUri.has_value() ? jni::make_jstring(startSoundUri.value()) : nullptr, endSoundUri.has_value() ? jni::make_jstring(endSoundUri.value()) : nullptr);
+  std::shared_ptr<Promise<VoiceInputResult>> JHybridVoiceSpec::startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri, std::optional<VoiceAudioEncoding> encoding) {
+    static const auto method = _javaPart->javaClassStatic()->getMethod<jni::local_ref<JPromise::javaobject>(jni::alias_ref<jni::JDouble> /* silenceThresholdMs */, jni::alias_ref<jni::JDouble> /* maxDurationMs */, jni::alias_ref<jni::JString> /* listeningText */, jni::alias_ref<JVariant_GlyphImage_AssetImage_RemoteImage> /* listeningImage */, jni::alias_ref<jni::JBoolean> /* listeningImageRepeats */, jni::alias_ref<jni::JBoolean> /* preferSpeechToText */, jni::alias_ref<JFunc_void_VoiceInputChunk::javaobject> /* onChunk */, jni::alias_ref<jni::JString> /* language */, jni::alias_ref<jni::JString> /* startSoundUri */, jni::alias_ref<jni::JString> /* endSoundUri */, jni::alias_ref<JVoiceAudioEncoding> /* encoding */)>("startVoiceInput_cxx");
+    auto __result = method(_javaPart, silenceThresholdMs.has_value() ? jni::JDouble::valueOf(silenceThresholdMs.value()) : nullptr, maxDurationMs.has_value() ? jni::JDouble::valueOf(maxDurationMs.value()) : nullptr, listeningText.has_value() ? jni::make_jstring(listeningText.value()) : nullptr, listeningImage.has_value() ? JVariant_GlyphImage_AssetImage_RemoteImage::fromCpp(listeningImage.value()) : nullptr, listeningImageRepeats.has_value() ? jni::JBoolean::valueOf(listeningImageRepeats.value()) : nullptr, preferSpeechToText.has_value() ? jni::JBoolean::valueOf(preferSpeechToText.value()) : nullptr, onChunk.has_value() ? JFunc_void_VoiceInputChunk_cxx::fromCpp(onChunk.value()) : nullptr, language.has_value() ? jni::make_jstring(language.value()) : nullptr, startSoundUri.has_value() ? jni::make_jstring(startSoundUri.value()) : nullptr, endSoundUri.has_value() ? jni::make_jstring(endSoundUri.value()) : nullptr, encoding.has_value() ? JVoiceAudioEncoding::fromCpp(encoding.value()) : nullptr);
     return [&]() {
       auto __promise = Promise<VoiceInputResult>::create();
       __result->cthis()->addOnResolvedListener([=](const jni::alias_ref<jni::JObject>& __boxedResult) {

package/nitrogen/generated/android/c++/JHybridVoiceSpec.hpp

@@ -56,7 +56,7 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay {
     // Methods
     bool hasVoiceInputPermission() override;
     std::shared_ptr<Promise<bool>> requestVoiceInputPermission() override;
-    std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri) override;
+    std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri, std::optional<VoiceAudioEncoding> encoding) override;
     void stopVoiceInput() override;
 
   private:

package/nitrogen/generated/android/c++/JVoiceAudioEncoding.hpp

@@ -0,0 +1,61 @@
+///
+/// JVoiceAudioEncoding.hpp
+/// This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+/// https://github.com/mrousavy/nitro
+/// Copyright © Marc Rousavy @ Margelo
+///
+
+#pragma once
+
+#include <fbjni/fbjni.h>
+#include "VoiceAudioEncoding.hpp"
+
+namespace margelo::nitro::swe::iternio::reactnativeautoplay {
+
+  using namespace facebook;
+
+  /**
+   * The C++ JNI bridge between the C++ enum "VoiceAudioEncoding" and the the Kotlin enum "VoiceAudioEncoding".
+   */
+  struct JVoiceAudioEncoding final: public jni::JavaClass<JVoiceAudioEncoding> {
+  public:
+    static constexpr auto kJavaDescriptor = "Lcom/margelo/nitro/swe/iternio/reactnativeautoplay/VoiceAudioEncoding;";
+
+  public:
+    /**
+     * Convert this Java/Kotlin-based enum to the C++ enum VoiceAudioEncoding.
+     */
+    [[maybe_unused]]
+    [[nodiscard]]
+    VoiceAudioEncoding toCpp() const {
+      static const auto clazz = javaClassStatic();
+      static const auto fieldOrdinal = clazz->getField<int>("value");
+      int ordinal = this->getFieldValue(fieldOrdinal);
+      return static_cast<VoiceAudioEncoding>(ordinal);
+    }
+
+  public:
+    /**
+     * Create a Java/Kotlin-based enum with the given C++ enum's value.
+     */
+    [[maybe_unused]]
+    static jni::alias_ref<JVoiceAudioEncoding> fromCpp(VoiceAudioEncoding value) {
+      static const auto clazz = javaClassStatic();
+      switch (value) {
+        case VoiceAudioEncoding::LINEAR16:
+          static const auto fieldLINEAR16 = clazz->getStaticField<JVoiceAudioEncoding>("LINEAR16");
+          return clazz->getStaticFieldValue(fieldLINEAR16);
+        case VoiceAudioEncoding::MULAW:
+          static const auto fieldMULAW = clazz->getStaticField<JVoiceAudioEncoding>("MULAW");
+          return clazz->getStaticFieldValue(fieldMULAW);
+        case VoiceAudioEncoding::ALAW:
+          static const auto fieldALAW = clazz->getStaticField<JVoiceAudioEncoding>("ALAW");
+          return clazz->getStaticFieldValue(fieldALAW);
+        default:
+          std::string stringValue = std::to_string(static_cast<int>(value));
+          throw std::invalid_argument("Invalid enum value (" + stringValue + "!");
+      }
+    }
+  };
+
+} // namespace margelo::nitro::swe::iternio::reactnativeautoplay

package/nitrogen/generated/android/kotlin/com/margelo/nitro/swe/iternio/reactnativeautoplay/HybridVoiceSpec.kt

@@ -37,12 +37,12 @@ abstract class HybridVoiceSpec: HybridObject() {
   @Keep
   abstract fun requestVoiceInputPermission(): Promise<Boolean>
   
-  abstract fun startVoiceInput(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Boolean?, preferSpeechToText: Boolean?, onChunk: ((chunk: VoiceInputChunk) -> Unit)?, language: String?, startSoundUri: String?, endSoundUri: String?): Promise<VoiceInputResult>
+  abstract fun startVoiceInput(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Boolean?, preferSpeechToText: Boolean?, onChunk: ((chunk: VoiceInputChunk) -> Unit)?, language: String?, startSoundUri: String?, endSoundUri: String?, encoding: VoiceAudioEncoding?): Promise<VoiceInputResult>
   
   @DoNotStrip
   @Keep
-  private fun startVoiceInput_cxx(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Boolean?, preferSpeechToText: Boolean?, onChunk: Func_void_VoiceInputChunk?, language: String?, startSoundUri: String?, endSoundUri: String?): Promise<VoiceInputResult> {
-    val __result = startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, listeningImage, listeningImageRepeats, preferSpeechToText, onChunk?.let { it }, language, startSoundUri, endSoundUri)
+  private fun startVoiceInput_cxx(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Boolean?, preferSpeechToText: Boolean?, onChunk: Func_void_VoiceInputChunk?, language: String?, startSoundUri: String?, endSoundUri: String?, encoding: VoiceAudioEncoding?): Promise<VoiceInputResult> {
+    val __result = startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, listeningImage, listeningImageRepeats, preferSpeechToText, onChunk?.let { it }, language, startSoundUri, endSoundUri, encoding)
     return __result
   }
   

package/nitrogen/generated/android/kotlin/com/margelo/nitro/swe/iternio/reactnativeautoplay/VoiceAudioEncoding.kt

@@ -0,0 +1,24 @@
+///
+/// VoiceAudioEncoding.kt
+/// This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+/// https://github.com/mrousavy/nitro
+/// Copyright © Marc Rousavy @ Margelo
+///
+
+package com.margelo.nitro.swe.iternio.reactnativeautoplay
+
+import androidx.annotation.Keep
+import com.facebook.proguard.annotations.DoNotStrip
+
+/**
+ * Represents the JavaScript enum/union "VoiceAudioEncoding".
+ */
+@DoNotStrip
+@Keep
+enum class VoiceAudioEncoding(@DoNotStrip @Keep val value: Int) {
+  LINEAR16(0),
+  MULAW(1),
+  ALAW(2);
+
+  companion object
+}

package/nitrogen/generated/ios/ReactNativeAutoPlay-Swift-Cxx-Bridge.hpp

@@ -128,6 +128,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class TurnTyp
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VisibilityState; }
 // Forward declaration of `VisibleTravelEstimate` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VisibleTravelEstimate; }
+// Forward declaration of `VoiceAudioEncoding` to properly resolve imports.
+namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VoiceAudioEncoding; }
 // Forward declaration of `VoiceInputChunk` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputChunk; }
 // Forward declaration of `VoiceInputResult` to properly resolve imports.
@@ -217,6 +219,7 @@ namespace ReactNativeAutoPlay { class HybridVoiceSpec_cxx; }
 #include "TurnType.hpp"
 #include "VisibilityState.hpp"
 #include "VisibleTravelEstimate.hpp"
+#include "VoiceAudioEncoding.hpp"
 #include "VoiceInputChunk.hpp"
 #include "VoiceInputResult.hpp"
 #include "ZoomEvent.hpp"
@@ -1680,6 +1683,21 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay::bridge::swift {
     return optional.value();
   }
   
+  // pragma MARK: std::optional<VoiceAudioEncoding>
+  /**
+   * Specialized version of `std::optional<VoiceAudioEncoding>`.
+   */
+  using std__optional_VoiceAudioEncoding_ = std::optional<VoiceAudioEncoding>;
+  inline std::optional<VoiceAudioEncoding> create_std__optional_VoiceAudioEncoding_(const VoiceAudioEncoding& value) noexcept {
+    return std::optional<VoiceAudioEncoding>(value);
+  }
+  inline bool has_value_std__optional_VoiceAudioEncoding_(const std::optional<VoiceAudioEncoding>& optional) noexcept {
+    return optional.has_value();
+  }
+  inline VoiceAudioEncoding get_std__optional_VoiceAudioEncoding_(const std::optional<VoiceAudioEncoding>& optional) noexcept {
+    return optional.value();
+  }
+  
   // pragma MARK: std::shared_ptr<HybridVoiceSpec>
   /**
    * Specialized version of `std::shared_ptr<HybridVoiceSpec>`.

package/nitrogen/generated/ios/ReactNativeAutoPlay-Swift-Cxx-Umbrella.hpp

@@ -150,6 +150,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class TurnTyp
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VisibilityState; }
 // Forward declaration of `VisibleTravelEstimate` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VisibleTravelEstimate; }
+// Forward declaration of `VoiceAudioEncoding` to properly resolve imports.
+namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VoiceAudioEncoding; }
 // Forward declaration of `VoiceInputChunk` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputChunk; }
 // Forward declaration of `VoiceInputResult` to properly resolve imports.
@@ -229,6 +231,7 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class ZoomEve
 #include "TurnType.hpp"
 #include "VisibilityState.hpp"
 #include "VisibleTravelEstimate.hpp"
+#include "VoiceAudioEncoding.hpp"
 #include "VoiceInputChunk.hpp"
 #include "VoiceInputResult.hpp"
 #include "ZoomEvent.hpp"

package/nitrogen/generated/ios/c++/HybridVoiceSpecSwift.hpp

@@ -26,6 +26,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct RemoteImage
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct NitroColor; }
 // Forward declaration of `VoiceInputChunk` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputChunk; }
+// Forward declaration of `VoiceAudioEncoding` to properly resolve imports.
+namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VoiceAudioEncoding; }
 
 #include <NitroModules/Promise.hpp>
 #include "VoiceInputResult.hpp"
@@ -40,6 +42,7 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputC
 #include "NitroColor.hpp"
 #include "VoiceInputChunk.hpp"
 #include <functional>
+#include "VoiceAudioEncoding.hpp"
 
 #include "ReactNativeAutoPlay-Swift-Cxx-Umbrella.hpp"
 
@@ -107,8 +110,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay {
       auto __value = std::move(__result.value());
       return __value;
     }
-    inline std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri) override {
-      auto __result = _swiftPart.startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, listeningImage, listeningImageRepeats, preferSpeechToText, onChunk, language, startSoundUri, endSoundUri);
+    inline std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri, std::optional<VoiceAudioEncoding> encoding) override {
+      auto __result = _swiftPart.startVoiceInput(silenceThresholdMs, maxDurationMs, listeningText, listeningImage, listeningImageRepeats, preferSpeechToText, onChunk, language, startSoundUri, endSoundUri, encoding);
       if (__result.hasError()) [[unlikely]] {
         std::rethrow_exception(__result.error());
       }

package/nitrogen/generated/ios/swift/HybridVoiceSpec.swift

@@ -15,7 +15,7 @@ public protocol HybridVoiceSpec_protocol: HybridObject {
   // Methods
   func hasVoiceInputPermission() throws -> Bool
   func requestVoiceInputPermission() throws -> Promise<Bool>
-  func startVoiceInput(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Bool?, preferSpeechToText: Bool?, onChunk: ((_ chunk: VoiceInputChunk) -> Void)?, language: String?, startSoundUri: String?, endSoundUri: String?) throws -> Promise<VoiceInputResult>
+  func startVoiceInput(silenceThresholdMs: Double?, maxDurationMs: Double?, listeningText: String?, listeningImage: Variant_GlyphImage_AssetImage_RemoteImage?, listeningImageRepeats: Bool?, preferSpeechToText: Bool?, onChunk: ((_ chunk: VoiceInputChunk) -> Void)?, language: String?, startSoundUri: String?, endSoundUri: String?, encoding: VoiceAudioEncoding?) throws -> Promise<VoiceInputResult>
   func stopVoiceInput() throws -> Void
 }
 

package/nitrogen/generated/ios/swift/HybridVoiceSpec_cxx.swift

@@ -156,7 +156,7 @@ open class HybridVoiceSpec_cxx {
   }
   
   @inline(__always)
-  public final func startVoiceInput(silenceThresholdMs: bridge.std__optional_double_, maxDurationMs: bridge.std__optional_double_, listeningText: bridge.std__optional_std__string_, listeningImage: bridge.std__optional_std__variant_GlyphImage__AssetImage__RemoteImage__, listeningImageRepeats: bridge.std__optional_bool_, preferSpeechToText: bridge.std__optional_bool_, onChunk: bridge.std__optional_std__function_void_const_VoiceInputChunk_____chunk______, language: bridge.std__optional_std__string_, startSoundUri: bridge.std__optional_std__string_, endSoundUri: bridge.std__optional_std__string_) -> bridge.Result_std__shared_ptr_Promise_VoiceInputResult___ {
+  public final func startVoiceInput(silenceThresholdMs: bridge.std__optional_double_, maxDurationMs: bridge.std__optional_double_, listeningText: bridge.std__optional_std__string_, listeningImage: bridge.std__optional_std__variant_GlyphImage__AssetImage__RemoteImage__, listeningImageRepeats: bridge.std__optional_bool_, preferSpeechToText: bridge.std__optional_bool_, onChunk: bridge.std__optional_std__function_void_const_VoiceInputChunk_____chunk______, language: bridge.std__optional_std__string_, startSoundUri: bridge.std__optional_std__string_, endSoundUri: bridge.std__optional_std__string_, encoding: bridge.std__optional_VoiceAudioEncoding_) -> bridge.Result_std__shared_ptr_Promise_VoiceInputResult___ {
     do {
       let __result = try self.__implementation.startVoiceInput(silenceThresholdMs: { () -> Double? in
         if bridge.has_value_std__optional_double_(silenceThresholdMs) {
@@ -248,7 +248,7 @@ open class HybridVoiceSpec_cxx {
         } else {
           return nil
         }
-      }())
+      }(), encoding: encoding.value)
       let __resultCpp = { () -> bridge.std__shared_ptr_Promise_VoiceInputResult__ in
         let __promise = bridge.create_std__shared_ptr_Promise_VoiceInputResult__()
         let __promiseHolder = bridge.wrap_std__shared_ptr_Promise_VoiceInputResult__(__promise)

package/nitrogen/generated/ios/swift/VoiceAudioEncoding.swift

@@ -0,0 +1,44 @@
+///
+/// VoiceAudioEncoding.swift
+/// This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+/// https://github.com/mrousavy/nitro
+/// Copyright © Marc Rousavy @ Margelo
+///
+
+/**
+ * Represents the JS union `VoiceAudioEncoding`, backed by a C++ enum.
+ */
+public typealias VoiceAudioEncoding = margelo.nitro.swe.iternio.reactnativeautoplay.VoiceAudioEncoding
+
+public extension VoiceAudioEncoding {
+  /**
+   * Get a VoiceAudioEncoding for the given String value, or
+   * return `nil` if the given value was invalid/unknown.
+   */
+  init?(fromString string: String) {
+    switch string {
+      case "LINEAR16":
+        self = .linear16
+      case "MULAW":
+        self = .mulaw
+      case "ALAW":
+        self = .alaw
+      default:
+        return nil
+    }
+  }
+
+  /**
+   * Get the String value this VoiceAudioEncoding represents.
+   */
+  var stringValue: String {
+    switch self {
+      case .linear16:
+        return "LINEAR16"
+      case .mulaw:
+        return "MULAW"
+      case .alaw:
+        return "ALAW"
+    }
+  }
+}

package/nitrogen/generated/shared/c++/HybridVoiceSpec.hpp

@@ -23,6 +23,8 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct AssetImage;
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct RemoteImage; }
 // Forward declaration of `VoiceInputChunk` to properly resolve imports.
 namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputChunk; }
+// Forward declaration of `VoiceAudioEncoding` to properly resolve imports.
+namespace margelo::nitro::swe::iternio::reactnativeautoplay { enum class VoiceAudioEncoding; }
 
 #include <NitroModules/Promise.hpp>
 #include "VoiceInputResult.hpp"
@@ -34,6 +36,7 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay { struct VoiceInputC
 #include <variant>
 #include "VoiceInputChunk.hpp"
 #include <functional>
+#include "VoiceAudioEncoding.hpp"
 
 namespace margelo::nitro::swe::iternio::reactnativeautoplay {
 
@@ -68,7 +71,7 @@ namespace margelo::nitro::swe::iternio::reactnativeautoplay {
       // Methods
       virtual bool hasVoiceInputPermission() = 0;
       virtual std::shared_ptr<Promise<bool>> requestVoiceInputPermission() = 0;
-      virtual std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri) = 0;
+      virtual std::shared_ptr<Promise<VoiceInputResult>> startVoiceInput(std::optional<double> silenceThresholdMs, std::optional<double> maxDurationMs, const std::optional<std::string>& listeningText, const std::optional<std::variant<GlyphImage, AssetImage, RemoteImage>>& listeningImage, std::optional<bool> listeningImageRepeats, std::optional<bool> preferSpeechToText, const std::optional<std::function<void(const VoiceInputChunk& /* chunk */)>>& onChunk, const std::optional<std::string>& language, const std::optional<std::string>& startSoundUri, const std::optional<std::string>& endSoundUri, std::optional<VoiceAudioEncoding> encoding) = 0;
       virtual void stopVoiceInput() = 0;
 
     protected:

package/nitrogen/generated/shared/c++/VoiceAudioEncoding.hpp

@@ -0,0 +1,80 @@
+///
+/// VoiceAudioEncoding.hpp
+/// This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+/// https://github.com/mrousavy/nitro
+/// Copyright © Marc Rousavy @ Margelo
+///
+
+#pragma once
+
+#if __has_include(<NitroModules/NitroHash.hpp>)
+#include <NitroModules/NitroHash.hpp>
+#else
+#error NitroModules cannot be found! Are you sure you installed NitroModules properly?
+#endif
+#if __has_include(<NitroModules/JSIConverter.hpp>)
+#include <NitroModules/JSIConverter.hpp>
+#else
+#error NitroModules cannot be found! Are you sure you installed NitroModules properly?
+#endif
+#if __has_include(<NitroModules/NitroDefines.hpp>)
+#include <NitroModules/NitroDefines.hpp>
+#else
+#error NitroModules cannot be found! Are you sure you installed NitroModules properly?
+#endif
+
+namespace margelo::nitro::swe::iternio::reactnativeautoplay {
+
+  /**
+   * An enum which can be represented as a JavaScript union (VoiceAudioEncoding).
+   */
+  enum class VoiceAudioEncoding {
+    LINEAR16      SWIFT_NAME(linear16) = 0,
+    MULAW      SWIFT_NAME(mulaw) = 1,
+    ALAW      SWIFT_NAME(alaw) = 2,
+  } CLOSED_ENUM;
+
+} // namespace margelo::nitro::swe::iternio::reactnativeautoplay
+
+namespace margelo::nitro {
+
+  // C++ VoiceAudioEncoding <> JS VoiceAudioEncoding (union)
+  template <>
+  struct JSIConverter<margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding> final {
+    static inline margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding fromJSI(jsi::Runtime& runtime, const jsi::Value& arg) {
+      std::string unionValue = JSIConverter<std::string>::fromJSI(runtime, arg);
+      switch (hashString(unionValue.c_str(), unionValue.size())) {
+        case hashString("LINEAR16"): return margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::LINEAR16;
+        case hashString("MULAW"): return margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::MULAW;
+        case hashString("ALAW"): return margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::ALAW;
+        default: [[unlikely]]
+          throw std::invalid_argument("Cannot convert \"" + unionValue + "\" to enum VoiceAudioEncoding - invalid value!");
+      }
+    }
+    static inline jsi::Value toJSI(jsi::Runtime& runtime, margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding arg) {
+      switch (arg) {
+        case margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::LINEAR16: return JSIConverter<std::string>::toJSI(runtime, "LINEAR16");
+        case margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::MULAW: return JSIConverter<std::string>::toJSI(runtime, "MULAW");
+        case margelo::nitro::swe::iternio::reactnativeautoplay::VoiceAudioEncoding::ALAW: return JSIConverter<std::string>::toJSI(runtime, "ALAW");
+        default: [[unlikely]]
+          throw std::invalid_argument("Cannot convert VoiceAudioEncoding to JS - invalid value: "
+                                    + std::to_string(static_cast<int>(arg)) + "!");
+      }
+    }
+    static inline bool canConvert(jsi::Runtime& runtime, const jsi::Value& value) {
+      if (!value.isString()) {
+        return false;
+      }
+      std::string unionValue = JSIConverter<std::string>::fromJSI(runtime, value);
+      switch (hashString(unionValue.c_str(), unionValue.size())) {
+        case hashString("LINEAR16"):
+        case hashString("MULAW"):
+        case hashString("ALAW"):
+          return true;
+        default:
+          return false;
+      }
+    }
+  };
+
+} // namespace margelo::nitro

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@iternio/react-native-auto-play",
-  "version": "0.5.10",
+  "version": "0.5.12",
   "description": "Android Auto and Apple CarPlay for react-native",
   "main": "lib/index",
   "module": "lib/index",

package/src/hybrid/HybridVoice.ts

@@ -24,6 +24,7 @@ const startVoiceInput: StartVoiceInput = async (options?: VoiceInputOptions) =>
     language,
     startSound,
     endSound,
+    encoding,
   } = options ?? {};
 
   const listeningImageRepeats =
@@ -42,7 +43,8 @@ const startVoiceInput: StartVoiceInput = async (options?: VoiceInputOptions) =>
     onChunk,
     language,
     startSoundUri,
-    endSoundUri
+    endSoundUri,
+    encoding
   );
 };
 
@@ -82,6 +84,7 @@ export const HybridVoice = {
    * @param preferSpeechToText  request STT transcription instead of raw PCM
    * @param onChunk             optional streaming callback
    * @param language            specify the language for the SpeechRecognizer, falls back to system language if not set
+   * @param encoding            PCM encoding for onChunk audio and the final result (default LINEAR16)
    */
   startVoiceInput,
   /**

package/src/specs/Voice.nitro.ts

@@ -1,5 +1,5 @@
 import type { HybridObject } from 'react-native-nitro-modules';
-import type { VoiceInputChunk, VoiceInputResult } from '../types/Voice';
+import type { VoiceAudioEncoding, VoiceInputChunk, VoiceInputResult } from '../types/Voice';
 import type { NitroImage } from '../utils/NitroImage';
 
 export interface Voice extends HybridObject<{ android: 'kotlin'; ios: 'swift' }> {
@@ -15,7 +15,8 @@ export interface Voice extends HybridObject<{ android: 'kotlin'; ios: 'swift' }>
     onChunk?: (chunk: VoiceInputChunk) => void,
     language?: string,
     startSoundUri?: string,
-    endSoundUri?: string
+    endSoundUri?: string,
+    encoding?: VoiceAudioEncoding
   ): Promise<VoiceInputResult>;
   stopVoiceInput(): void;
 }

package/src/types/Voice.ts

@@ -1,5 +1,10 @@
 import type { VoiceInputImage } from './Image';
 
+/** PCM sample encoding for the `audio` payload of chunks and the final result.
+ * `LINEAR16` is raw 16-bit signed PCM
+ * `MULAW`/`ALAW` are G.711 8-bit companded encodings, halving payload size */
+export type VoiceAudioEncoding = 'LINEAR16' | 'MULAW' | 'ALAW';
+
 export interface VoiceInputChunk {
   partial?: string;
   audio?: ArrayBuffer;
@@ -21,6 +26,8 @@ export interface VoiceInputOptions {
   preferSpeechToText?: boolean;
   onChunk?: (chunk: VoiceInputChunk) => void;
   language?: string;
+  /** PCM encoding for onChunk audio and the final result. Defaults to LINEAR16. */
+  encoding?: VoiceAudioEncoding;
   /** Sound played just before recording starts. Pass a Metro asset: `require('./beep_start.wav')`. */
   startSound?: number;
   /** Sound played just after recording stops. Pass a Metro asset: `require('./beep_end.wav')`. */