@gmessier/nitro-speech 0.4.0 → 0.4.2

package/README.md

@@ -13,7 +13,7 @@
 
 #### Key Features:
 
-- ⚡ Built on Nitro Modules for zero-overhead native bridging
+- ⚡ Built with Nitro Modules for low-overhead native binding
 - 🌎 Supports 60+ languages 
 - 🍎 The only library that uses new `SpeechAnalyzer` with `SpeechTranscriber` or `DictationTranscriber` API for iOS 26+ (with fallback to legacy `SFSpeechRecognition` for older versions)
 - ⏱️ Timer for silence
@@ -21,16 +21,16 @@
   - Callback `onAutoFinishProgress` fires periodically with interval
   - Configurable interval `autoFinishProgressIntervalMs` value (default: 1 sec)
   - Method `updateConfig` with `autoFinishRecognitionMs` and `autoFinishProgressIntervalMs`
-  allows to change value on the fly
+  allows changing the value on the fly
   - Method `resetAutoFinishTime` resets the Timer to the threshold
   - Method `addAutoFinishTime` adds ms once without changing threshold
   - Configurable volume-based sensitivity `resetAutoFinishVoiceSensitivity` for the timer from 0 to 1
 - 🎤 Rich user voice input management 
-  - Hook `useVoiceInputVolume()` for `raw` or `smoothed` normalized 
-  volume level from 0 to 1 -> easy to use for UI animations;
+  - Hook `useVoiceInputVolume()` for `raw` or `smoothed` normalized volume level from 0 to 1 -> easy to use for UI animations;
   And `db` as human-friendly value
   - Flexible callback `onVolumeChange` for custom behavior
-- 🧩 Lifecycle methods: `prewarm` | `updateConfig` | `getIsActive` | `getSupportedLocalesIOS`
+  - Static method `getVoiceInputVolume()`
+- 🧩 Lifecycle methods: `prewarm` | `updateConfig` | `getIsActive`
 - 👆 Configurable Haptic Feedback on start and finish
 - 🎚️ Speech-quality configurations:
   - Result is grouped by speech segments into Batches.
@@ -53,6 +53,7 @@
   - [Cross-component control: RecognizerRef](#cross-component-control-recognizerref)
   - [Multithreading (react-native-worklets)](#multithreading-react-native-worklets)
   - [Voice input volume](#voice-input-volume)
+  - [useRecognizerIsActive](#userecognizerisactive)
   - [Unsafe: SpeechRecognizer](#unsafe-speechrecognizer)
 - [Requirements](#requirements)
 - [Compatibility](#compatibility)
@@ -123,11 +124,11 @@ Both permissions are required for speech recognition to work on iOS.
 | **Auto-finish progress** | Callback `onAutoFinishProgress` with countdown until auto-stop | ✅ | ✅ |
 | **Add Auto-finish Time** | Adds time to the auto finish timer once without changing the timer threshold | ✅ | ✅ |
 | **Reset Auto-finish Time** | Resets the Timer to the threshold | ✅ | ✅ |
-| **Voice input volume** | Hook `useVoiceInputVolume` and `onVolumeChange` callback | ✅ | ✅ |
+| **Voice input volume** | `useVoiceInputVolume`, `getVoiceInputVolume()`, `onVolumeChange` | ✅ | ✅ |
 | **Reset Auto-finish Sensitivity** | The voice detector sensitivity to reset the Auto-finish time | ✅ | ✅ |
 | **Prewarm** | Prepares resources, downloads assets, confirms locale availability | ✅ | ✅ |
-| **Update config** | Static method `updateConfig` allows update config on the fly | ✅ | ✅ |
-| **isActive** | Static method `getIsActive()` | ✅ | ✅ |
+| **Update config** | Static method `updateConfig` allows updating the config on the fly | ✅ | ✅ |
+| **Is Active** | Static method `getIsActive()` | ✅ | ✅ |
 | **Haptic feedback** | Haptic feedback on recording start/stop | ✅ | ✅ |
 | **Permission handling** | Dedicated `onPermissionDenied` callback | ✅ | ✅ |
 | **Background handling** | Stop when app loses focus/goes to background | ✅ | ✅ |
@@ -154,7 +155,8 @@ Because of that, treat it as a **single session owner** setup hook: use it once
 import { useRecognizer } from '@gmessier/nitro-speech';
 
 function MyComponent() {
-  const { 
+  const {
+    prewarm,
     startListening, 
     stopListening, 
     resetAutoFinishTime, 
@@ -162,6 +164,7 @@ function MyComponent() {
     updateConfig,
     getSupportedLocalesIOS,
     getIsActive,
+    getVoiceInputVolume,
   } = useRecognizer({
     onReadyForSpeech: () => {
       console.log('Listening...');
@@ -259,6 +262,7 @@ If you need to call recognizer methods from other components without prop drilli
 ```typescript
 import { RecognizerRef } from '@gmessier/nitro-speech';
 
+RecognizerRef.prewarm({ locale: 'en-US' });
 RecognizerRef.startListening({ locale: 'en-US' });
 RecognizerRef.addAutoFinishTime(5000);
 RecognizerRef.resetAutoFinishTime();
@@ -271,6 +275,7 @@ RecognizerRef.updateConfig(
   true
 );
 RecognizerRef.getIsActive();
+RecognizerRef.getVoiceInputVolume();
 RecognizerRef.stopListening();
 // iOS only
 RecognizerRef.getSupportedLocalesIOS();
@@ -321,8 +326,6 @@ function VoiceMeter() {
 As a better alternative you can control volume via SharedValue and apply it only on UI thread with Reanimated.
 This way you will avoid re-renders since the volume will be stored on UI thread
 
-Warning: this approach will disable the built-in `useVoiceInputVolume` hook.
-
 ```typescript
 function VoiceMeter() {
   const sharedVolume = useSharedValue(0)
@@ -341,17 +344,35 @@ function VoiceMeter() {
 }
 ```
 
+### useRecognizerIsActive
+
+```typescript
+import { useRecognizerIsActive } from '@gmessier/nitro-speech';
+
+function MyComponent() {
+  const isActive = useRecognizerIsActive();
+  return <Text>{isActive ? 'Listening...' : 'Not listening'}</Text>;
+}
+```
 
 ### Unsafe: SpeechRecognizer
 
 `SpeechRecognizer` is the hybrid object. It gives direct access to callbacks and control methods, but it is unsafe to orchestrate the full session directly from it.
 
+**Warning**: Since it reflects the original hybrid object, its API may change in the future.
+
 ```typescript
-import { SpeechRecognizer, speechRecognizerVolumeChangeHandler } from '@gmessier/nitro-speech';
+import { 
+  SpeechRecognizer, 
+  speechRecognizerVolumeChangeHandler,
+  speechRecognizerActiveStateHandler,
+} from '@gmessier/nitro-speech';
 
 // Set up callbacks
 SpeechRecognizer.onReadyForSpeech = () => {
   console.log('Listening...');
+  // Add speechRecognizerActiveStateHandler to enable useRecognizerIsActive hook manually
+  speechRecognizerActiveStateHandler(true);
 };
 
 SpeechRecognizer.onResult = (textBatches) => {
@@ -360,6 +381,8 @@ SpeechRecognizer.onResult = (textBatches) => {
 
 SpeechRecognizer.onRecordingStopped = () => {
   console.log('Stopped');
+  // Add speechRecognizerActiveStateHandler to enable useRecognizerIsActive hook manually
+  speechRecognizerActiveStateHandler(false);
 };
 
 SpeechRecognizer.onAutoFinishProgress = (timeLeftMs) => {
@@ -376,10 +399,27 @@ SpeechRecognizer.onPermissionDenied = () => {
 
 SpeechRecognizer.onVolumeChange = (volume) => {
   console.log('new volume: ', volume);
+  // Add speechRecognizerVolumeChangeHandler to enable useVoiceInputVolume hook manually
+  speechRecognizerVolumeChangeHandler(volume);
 };
-// OR use speechRecognizerVolumeChangeHandler to enable useVoiceInputVolume hook manually
-SpeechRecognizer.onVolumeChange = speechRecognizerVolumeChangeHandler
 
+// Prepare resources, download assets, confirms locale availability
+SpeechRecognizer.prewarm({
+  locale: 'en-US',
+  // ... your config to prepare
+});
+// OR `await` if you want to react to the success
+await SpeechRecognizer.prewarm({
+  locale: 'en-US',
+  // ... your config to prepare
+});
+// OR from worklet (only sync)
+scheduleOnRuntime(workletRuntime, () => {
+  SpeechRecognizer.prewarm({
+    locale: 'en-US',
+    // ... your config to prepare
+  });
+});
 
 // Start listening
 SpeechRecognizer.startListening({
@@ -415,7 +455,7 @@ The `SpeechRecognizer.dispose()` method is **NOT SAFE** and should rarely be use
 
 ## Compatibility
 
-Latest versions of `@gmessier/nitro-speech` requires [react-native-nitro-modules 0.35.0 or higher](https://github.com/mrousavy/nitro/releases/tag/v0.35.0).
+Latest versions of `@gmessier/nitro-speech` require [react-native-nitro-modules 0.35.0 or higher](https://github.com/mrousavy/nitro/releases/tag/v0.35.0).
 
 
 | Compatibility                          | Supported versions                |
@@ -427,7 +467,7 @@ Latest versions of `@gmessier/nitro-speech` requires [react-native-nitro-modules
 
 ### Android Gradle sync issues
 
-If you're having issues with Android Gradle sync, try running the prebuild for the library, that causes the issue:
+If you're having issues with Android Gradle sync, try running the prebuild for the library that causes the issue:
 
 e.g. failed in `react-native-nitro-modules`:
 

package/android/src/main/java/com/margelo/nitro/nitrospeech/recognizer/AutoStopper.kt

@@ -2,7 +2,6 @@ package com.margelo.nitro.nitrospeech.recognizer
 
 import android.os.Handler
 import android.os.Looper
-import android.util.Log
 import kotlin.math.max
 
 class AutoStopper(
@@ -12,12 +11,13 @@ class AutoStopper(
     val onTimeout: () -> Unit,
 ) {
     companion object {
-        private const val TAG = "HybridRecognizer"
         private const val DEFAULT_SILENCE_THRESHOLD_MS = 8000.0
         private const val DEFAULT_PROGRESS_INTERVAL_MS = 1000.0
         private const val MIN_PROGRESS_INTERVAL_MS = 50.0
     }
 
+    private val logger = Logger(disable = false)
+
     private var silenceThresholdMs: Double = clampMs(silenceThresholdMs ?: DEFAULT_SILENCE_THRESHOLD_MS)
     private var progressIntervalMs: Double = clampMs(progressIntervalMs ?: DEFAULT_PROGRESS_INTERVAL_MS)
 
@@ -31,7 +31,7 @@ class AutoStopper(
     private val tickRunnable = Runnable { tick() }
 
     fun resetTimer() {
-        Log.d(TAG, "resetTimer | isStopped: $isStopped | ms: ${System.currentTimeMillis()}")
+        logger.log("resetTimer | isStopped: $isStopped | ms: ${System.currentTimeMillis()}")
         handler.removeCallbacks(tickRunnable)
         isTimerScheduled = false
         if (isStopped) return
@@ -55,7 +55,7 @@ class AutoStopper(
 
     fun addMsOnce(extraMs: Double) {
         if (isStopped || !extraMs.isFinite()) return
-        Log.d(TAG, "addMsOnce | extraMs: $extraMs")
+        logger.log("addMsOnce | extraMs: $extraMs")
         timeLeftMs += extraMs
         didTimeout = false
         if (timeLeftMs > 0 && isTimerScheduled) {
@@ -65,7 +65,7 @@ class AutoStopper(
 
     fun updateProgressInterval(newIntervalMs: Double) {
         if (isStopped) return
-        Log.d(TAG, "updateProgressInterval | newIntervalMs: $newIntervalMs")
+        logger.log("updateProgressInterval | newIntervalMs: $newIntervalMs")
         progressIntervalMs = clampMs(newIntervalMs)
         if (isTimerScheduled) {
             scheduleNextTickLocked()
@@ -83,7 +83,7 @@ class AutoStopper(
         if (isStopped || didTimeout) return
         timeLeftMs -= progressIntervalMs
         if (timeLeftMs > 0) {
-            Log.d(TAG, "onProgress | timeLeftMs: $timeLeftMs")
+            logger.log("onProgress | timeLeftMs: $timeLeftMs")
             onProgress(timeLeftMs)
             scheduleNextTickLocked()
             return
@@ -92,7 +92,7 @@ class AutoStopper(
         didTimeout = true
         handler.removeCallbacks(tickRunnable)
         isTimerScheduled = false
-        Log.d(TAG, "onTimeout | ms: ${System.currentTimeMillis()}")
+        logger.log("onTimeout | ms: ${System.currentTimeMillis()}")
         onTimeout()
     }
 

package/android/src/main/java/com/margelo/nitro/nitrospeech/recognizer/HybridRecognizer.kt

@@ -7,7 +7,6 @@ import android.os.Handler
 import android.os.Looper
 import android.speech.RecognizerIntent
 import android.speech.SpeechRecognizer
-import android.util.Log
 import androidx.annotation.Keep
 import com.facebook.proguard.annotations.DoNotStrip
 import com.margelo.nitro.NitroModules
@@ -21,12 +20,14 @@ import com.margelo.nitro.nitrospeech.VolumeChangeEvent
 @Keep
 class HybridRecognizer: HybridRecognizerSpec() {
   companion object {
-    private const val TAG = "HybridRecognizer"
     private const val POST_RECOGNITION_DELAY = 250L
   }
 
+  private val logger = Logger(disable = false)
+
   private var isActive: Boolean = false
   private var config: SpeechRecognitionConfig? = null
+  private var volumeChangeEvent: VolumeChangeEvent = VolumeChangeEvent(0.0,0.0,null)
   private var autoStopper: AutoStopper? = null
   private var speechRecognizer: SpeechRecognizer? = null
   private val mainHandler = Handler(Looper.getMainLooper())
@@ -51,7 +52,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
   @DoNotStrip
   @Keep
   override fun startListening(params: SpeechRecognitionConfig?) {
-    Log.d(TAG, "startListening: $params")
+    logger.log("startListening: $params")
     if (isActive) {
       onFinishRecognition(
         null,
@@ -94,7 +95,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
   @DoNotStrip
   @Keep
   override fun stopListening() {
-    Log.d(TAG, "stopListening called")
+    logger.log("stopListening called")
     if (!isActive) return
     onFinishRecognition(null, null, true)
     mainHandler.postDelayed({
@@ -117,7 +118,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
   @DoNotStrip
   @Keep
   override fun addAutoFinishTime(additionalTimeMs: Double?) {
-    Log.d(TAG, "addAutoFinishTime")
+    logger.log("addAutoFinishTime")
     if (!isActive) return
 
     if (additionalTimeMs != null) {
@@ -134,7 +135,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
     newConfig: MutableSpeechRecognitionConfig?,
     resetAutoFinishTime: Boolean?
   ) {
-    Log.d(TAG, "updateConfig $newConfig",)
+    logger.log("updateConfig $newConfig",)
     if (!isActive) return
 
     val newTimeMs = if (newConfig?.autoFinishRecognitionMs != null) newConfig.autoFinishRecognitionMs else config?.autoFinishRecognitionMs
@@ -177,6 +178,12 @@ class HybridRecognizer: HybridRecognizerSpec() {
     return isActive
   }
 
+  @DoNotStrip
+  @Keep
+  override fun getVoiceInputVolume(): VolumeChangeEvent {
+    return volumeChangeEvent
+  }
+
   @DoNotStrip
   @Keep
   override fun getSupportedLocalesIOS(): Array<String> {
@@ -204,12 +211,14 @@ class HybridRecognizer: HybridRecognizerSpec() {
             }
         )
         val recognitionListenerSession = RecognitionListenerSession(
-          autoStopper,
-          config,
-          onVolumeChange
-        ) { result: ArrayList<String>?, errorMessage: String?, recordingStopped: Boolean ->
-          onFinishRecognition(result, errorMessage, recordingStopped)
-        }
+            autoStopper,
+            config,
+            fireVolumeChangeEvent = { event -> fireVolumeChangeEvent(event) },
+            onFinishRecognition = { result, errorMessage, recordingStopped ->
+              onFinishRecognition(result, errorMessage, recordingStopped)
+            }
+        )
+
         speechRecognizer?.setRecognitionListener(recognitionListenerSession.createRecognitionListener())
 
         val languageModel = if (config?.androidUseWebSearchModel == true) RecognizerIntent.LANGUAGE_MODEL_WEB_SEARCH else RecognizerIntent.LANGUAGE_MODEL_FREE_FORM
@@ -262,7 +271,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
 
   private fun cleanup() {
     try {
-      Log.d(TAG, "cleanup called")
+      logger.log("cleanup called")
       autoStopper?.stop()
       autoStopper = null
       speechRecognizer?.stopListening()
@@ -270,7 +279,7 @@ class HybridRecognizer: HybridRecognizerSpec() {
       speechRecognizer = null
       isActive = false
       // Reset voice meter in JS consumers after stop/error cleanup.
-      onVolumeChange?.invoke(VolumeChangeEvent(0.0,0.0,null))
+      fireVolumeChangeEvent(VolumeChangeEvent(0.0,0.0,null))
     } catch (e: Exception) {
       onFinishRecognition(
         null,
@@ -291,4 +300,10 @@ class HybridRecognizer: HybridRecognizerSpec() {
       onResult?.invoke(result.toTypedArray())
     }
   }
+
+  private fun fireVolumeChangeEvent(event: VolumeChangeEvent) {
+    logger.log("fireVolumeChangeEvent ${event}")
+    volumeChangeEvent = event
+    onVolumeChange?.invoke(event)
+  }
 }

package/android/src/main/java/com/margelo/nitro/nitrospeech/recognizer/Logger.kt

@@ -0,0 +1,16 @@
+package com.margelo.nitro.nitrospeech.recognizer
+
+import android.util.Log
+
+class Logger (
+  private val disable: Boolean
+) {
+  private val isLogging = false
+  companion object {
+    private const val TAG = "HybridRecognizer"
+  }
+  fun log(message: String) {
+    if (disable || !isLogging) return
+    Log.d(TAG, message)
+  }
+}

package/android/src/main/java/com/margelo/nitro/nitrospeech/recognizer/RecognitionListenerSession.kt

@@ -3,7 +3,6 @@ package com.margelo.nitro.nitrospeech.recognizer
 import android.os.Bundle
 import android.speech.RecognitionListener
 import android.speech.SpeechRecognizer
-import android.util.Log
 import com.margelo.nitro.nitrospeech.SpeechRecognitionConfig
 import com.margelo.nitro.nitrospeech.VolumeChangeEvent
 import kotlin.math.max
@@ -12,11 +11,11 @@ import kotlin.math.roundToInt
 class RecognitionListenerSession (
     private val autoStopper: AutoStopper?,
     private val config: SpeechRecognitionConfig?,
-    private val onVolumeChange: ((event: VolumeChangeEvent) -> Unit)?,
+    private val fireVolumeChangeEvent: (event: VolumeChangeEvent) -> Unit,
     private val onFinishRecognition: (result: ArrayList<String>?, errorMessage: String?, recordingStopped: Boolean) -> Unit,
 ) {
+    private val logger = Logger(disable = false)
     companion object {
-        private const val TAG = "HybridRecognizer"
         private const val SPEECH_LEVEL_THRESHOLD = 0.35
         private const val FLOOR_RISE_ALPHA = 0.01f
         private const val FLOOR_FALL_ALPHA = 0.20f
@@ -40,11 +39,11 @@ class RecognitionListenerSession (
             override fun onBeginningOfSpeech() {}
             override fun onRmsChanged(rmsdB: Float) {
                 val volumeEvent = getVolume(rmsdB)
-                onVolumeChange?.invoke(volumeEvent)
+                fireVolumeChangeEvent(volumeEvent)
                 val threshold =
                     config?.resetAutoFinishVoiceSensitivity?.coerceIn(0.0, 1.0)
                         ?: SPEECH_LEVEL_THRESHOLD.toDouble()
-                Log.d(TAG, "onRmsChanged: ${volumeEvent}")
+                // logger.log("onRmsChanged: ${volumeEvent}")
                 if (volumeEvent.rawVolume > threshold) {
                     autoStopper?.resetTimer()
                 }
@@ -75,7 +74,7 @@ class RecognitionListenerSession (
             }
 
             override fun onResults(results: Bundle?) {
-                Log.d(TAG, "onResults: $resultBatches")
+                logger.log("onResults: $resultBatches")
                 onFinishRecognition(resultBatches, null, true)
                 autoStopper?.stop()
                 autoStopper?.onTimeout()
@@ -85,26 +84,26 @@ class RecognitionListenerSession (
                 val matches = partialResults?.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION)
 
                 if (matches.isNullOrEmpty() || matches[0] == "") {
-                    Log.d(TAG, "onPartialResults[0], skip, NO RECOGNIZE")
+                    logger.log("onPartialResults[0], skip, NO RECOGNIZE")
                     return
                 }
 
                 autoStopper?.resetTimer()
-                Log.d(TAG, "onPartialResults[0], add ${matches[0]}")
+                logger.log("onPartialResults[0], add ${matches[0]}")
                 var currentBatches = resultBatches
                 if (currentBatches.isNullOrEmpty()) {
-                    Log.d(TAG, "onPartialResults[1], NO BATCHES YET | add first")
+                    logger.log("onPartialResults[1], NO BATCHES YET | add first")
                     currentBatches = arrayListOf(matches[0])
                 } else {
-                    Log.d(TAG, "onPartialResults[1], current batches $currentBatches")
+                    logger.log("onPartialResults[1], current batches $currentBatches")
                     val prevBatchLength = currentBatches[currentBatches.lastIndex].length
                     val match = if (config?.disableRepeatingFilter == true) matches[0] else repeatingFilter(matches[0])
                     val matchLength = match.length
                     if (config?.androidDisableBatchHandling == true || matchLength + 3 < prevBatchLength) {
-                        Log.d(TAG, "onPartialResults[2], append new batch")
+                        logger.log("onPartialResults[2], append new batch")
                         currentBatches.add(match)
                     } else {
-                        Log.d(TAG, "onPartialResults[2], update batch, replace #${currentBatches.lastIndex}")
+                        logger.log("onPartialResults[2], update batch, replace #${currentBatches.lastIndex}")
                         currentBatches[currentBatches.lastIndex] = match
                     }
                 }

package/ios/Audio/AudioLevelTracker.swift

@@ -16,33 +16,18 @@ final class AudioLevelTracker {
     private static let meterRelease: Float = 0.08
     private static let defaultAutoStopResetThreshold: Double = 0.4
 
-    private var autoStopResetThreshold: Double
     private var smoothedLevel: Float = 0
+
+    var currentSample: AudioLevelSample?
     
-    init(resetAutoFinishVoiceSensitivity: Double?) {
-        if let resetAutoFinishVoiceSensitivity {
-            // Clamp value between 0 and 1
-            self.autoStopResetThreshold = max(0, min(1, resetAutoFinishVoiceSensitivity))
-        } else {
-            self.autoStopResetThreshold = Self.defaultAutoStopResetThreshold
-        }
-    }
-    
-    func updateResetAutoFinishVoiceSensitivity(newValue: Double?) {
-        if let newValue {
-            // Clamp value between 0 and 1
-            self.autoStopResetThreshold = max(0, min(1, newValue))
-        } else {
-            self.autoStopResetThreshold = Self.defaultAutoStopResetThreshold
-        }
-    }
+    private let lg = Lg(prefix: "RecognizerEngine")
 
     func reset() {
         smoothedLevel = 0
-        self.autoStopResetThreshold = Self.defaultAutoStopResetThreshold
+        currentSample = nil
     }
 
-    func process(_ buffer: AVAudioPCMBuffer) -> AudioLevelSample? {
+    func process(_ buffer: AVAudioPCMBuffer,_ autoStopResetThreshold: Double? = nil) -> AudioLevelSample? {
         guard let samples = buffer.floatChannelData?[0] else { return nil }
 
         let frameCount = Int(buffer.frameLength)
@@ -56,11 +41,20 @@ final class AudioLevelTracker {
         let coeff = normalized > smoothedLevel ? Self.meterAttack : Self.meterRelease
         smoothedLevel += coeff * (normalized - smoothedLevel)
         
-        return AudioLevelSample(
+        var threshold = Self.defaultAutoStopResetThreshold
+        if let autoStopResetThreshold {
+            threshold = max(0, min(1, autoStopResetThreshold))
+        }
+
+        currentSample = AudioLevelSample(
             smoothed: Double(smoothedLevel * 1_000_000).rounded() / 1_000_000,
             raw: Double(normalized * 1_000_000).rounded() / 1_000_000,
             db: Double(db * 1_000).rounded() / 1_000,
-            resetTimer: Double(normalized) >= self.autoStopResetThreshold
+            resetTimer: Double(normalized) >= threshold
         )
+        
+        lg.log("[AudioLevelTracker.process] autoStopResetThreshold: \(threshold)")
+        
+        return currentSample
     }
 }

package/ios/Engines/RecognizerEngine.swift

@@ -18,7 +18,7 @@ class RecognizerEngine {
     var hardwareFormat: AVAudioFormat?
     weak var recognizerDelegate: RecognizerDelegate?
     
-    private let audioLevelTracker: AudioLevelTracker
+    private let audioLevelTracker = AudioLevelTracker()
     private var appStateObserver: AppStateObserver?
     private var audioEngine: AVAudioEngine?
     private var autoStopper: AutoStopper?
@@ -29,9 +29,6 @@ class RecognizerEngine {
     init(locale: Locale, delegate: RecognizerDelegate) {
         self.locale = locale
         self.recognizerDelegate = delegate
-        self.audioLevelTracker = AudioLevelTracker(
-            resetAutoFinishVoiceSensitivity: delegate.config?.resetAutoFinishVoiceSensitivity
-        )
     }
     
     // MARK: - Recognizer Methods
@@ -84,7 +81,10 @@ class RecognizerEngine {
             format: hardwareFormat
         ) { [weak self] buffer, _ in
             guard let self, let recognizerDelegate = self.recognizerDelegate else { return }
-            if let sample = self.audioLevelTracker.process(buffer) {
+            if let sample = self.audioLevelTracker.process(
+                buffer,
+                recognizerDelegate.config?.resetAutoFinishVoiceSensitivity
+            ) {
                 // Send buffer volume data
                 recognizerDelegate.volumeChange(
                     event:
@@ -148,13 +148,7 @@ class RecognizerEngine {
                 from: "updateSession"
             )
         }
-        // Update AutoFinish reset voice sensitivity interval
-        if let newSensitivity = newConfig?.resetAutoFinishVoiceSensitivity, 
-            newSensitivity != currentConfig?.resetAutoFinishVoiceSensitivity {
-            audioLevelTracker.updateResetAutoFinishVoiceSensitivity(
-                newValue: newSensitivity
-            )
-        }
+        
         if let addMsToTimer {
             // Add time to the timer once
             autoStopper?.addMsOnce(
@@ -168,7 +162,16 @@ class RecognizerEngine {
         // Only update new non-nil values in the config
         recognizerDelegate.softlyUpdateConfig(newConfig: newConfig)
     }
-    
+
+    func getVoiceInputVolume() -> VolumeChangeEvent? {
+        guard let currentSample = audioLevelTracker.currentSample else { return nil }
+        return VolumeChangeEvent(
+            smoothedVolume: currentSample.smoothed,
+            rawVolume: currentSample.raw,
+            db: currentSample.db
+        )
+    }
+
     func cleanup(from: String) {
         lg.log("[cleanup]: \(from)")
         let wasActive = isActive

package/ios/HybridRecognizer.swift

@@ -68,6 +68,14 @@ class HybridRecognizer: HybridRecognizerSpec  {
     func getIsActive() -> Bool {
         engine?.isActive ?? false
     }
+
+    func getVoiceInputVolume() -> VolumeChangeEvent {
+        return engine?.getVoiceInputVolume() ?? VolumeChangeEvent(
+            smoothedVolume: 0,
+            rawVolume: 0,
+            db: nil
+        )
+    }
     
     func getSupportedLocalesIOS() -> [String] {
         return self.coordinator.getSupportedLocales()

package/ios/Shared/AutoStopper.swift

@@ -5,7 +5,7 @@ final class AutoStopper {
     private static let defaultProgressIntervalMs = 1000.0
     private static let minProgressIntervalMs = 50.0
 
-    private let lg = Lg(prefix: "AutoStopper", disable: true)
+    private let lg = Lg(prefix: "AutoStopper", disable: false)
     
     private let queue = DispatchQueue(label: "com.margelo.nitrospeech.autostopper")
 

package/lib/Recognizer/RecognizerRef.d.ts

@@ -1,5 +1,7 @@
 import type { RecognizerMethods } from './types';
 /**
  * Safe cross-component reference to the Speech Recognizer methods.
+ *
+ * All methods support worklets and UI thread calls
  */
 export declare const RecognizerRef: RecognizerMethods;

package/lib/Recognizer/RecognizerRef.js

@@ -1,13 +1,17 @@
-import { recognizerAddAutoFinishTime, recognizerGetSupportedLocalesIOS, recognizerGetIsActive, recognizerResetAutoFinishTime, recognizerStartListening, recognizerStopListening, recognizerUpdateConfig, } from './methods';
+import { recognizerAddAutoFinishTime, recognizerGetSupportedLocalesIOS, recognizerGetIsActive, recognizerResetAutoFinishTime, recognizerStartListening, recognizerStopListening, recognizerUpdateConfig, recognizerGetVoiceInputVolume, recognizerPrewarm, } from './methods';
 /**
  * Safe cross-component reference to the Speech Recognizer methods.
+ *
+ * All methods support worklets and UI thread calls
  */
 export const RecognizerRef = {
+    prewarm: recognizerPrewarm,
     startListening: recognizerStartListening,
     stopListening: recognizerStopListening,
     resetAutoFinishTime: recognizerResetAutoFinishTime,
     addAutoFinishTime: recognizerAddAutoFinishTime,
     updateConfig: recognizerUpdateConfig,
     getIsActive: recognizerGetIsActive,
+    getVoiceInputVolume: recognizerGetVoiceInputVolume,
     getSupportedLocalesIOS: recognizerGetSupportedLocalesIOS,
 };

package/lib/Recognizer/SpeechRecognizer.d.ts

@@ -3,6 +3,7 @@
  *
  * Direct access to the all Speech Recognizer methods and callbacks.
  *
- * @note unsafe, might lead to race conditions
+ * @note Unsafe, might lead to race conditions
+ * @warning Since it reflects the original hybrid object, its API may change in the future.
  */
 export declare const SpeechRecognizer: import("./types").RecognizerSpec;

package/lib/Recognizer/SpeechRecognizer.js

@@ -4,6 +4,7 @@ import { NitroSpeech } from '../NitroSpeech';
  *
  * Direct access to the all Speech Recognizer methods and callbacks.
  *
- * @note unsafe, might lead to race conditions
+ * @note Unsafe, might lead to race conditions
+ * @warning Since it reflects the original hybrid object, its API may change in the future.
  */
 export const SpeechRecognizer = NitroSpeech.recognizer;