whisper.rn 0.4.1 → 0.4.2

package/android/src/main/java/com/rnwhisper/RNWhisper.java

@@ -415,7 +415,8 @@ public class RNWhisper implements LifecycleEventListener {
       @Override
       protected WritableArray doInBackground(Void... voids) {
         try {
-          return vadContext.detectSpeech(audioDataBase64, options);
+          float[] audioData = AudioUtils.decodePcmData(audioDataBase64);
+          return vadContext.detectSpeechWithAudioData(audioData, audioData.length, options);
         } catch (Exception e) {
           exception = e;
           return null;
@@ -468,7 +469,7 @@ public class RNWhisper implements LifecycleEventListener {
             throw new Exception("Failed to load audio file: " + filePathOrBase64);
           }
 
-          return vadContext.detectSpeechWithAudioData(audioData, options);
+          return vadContext.detectSpeechWithAudioData(audioData, audioData.length, options);
         } catch (Exception e) {
           exception = e;
           return null;

package/android/src/main/java/com/rnwhisper/WhisperVadContext.java

@@ -25,70 +25,14 @@ public class WhisperVadContext {
         this.reactContext = reactContext;
     }
 
-    public WritableArray detectSpeech(String audioDataBase64, ReadableMap options) throws Exception {
+    public WritableArray detectSpeechWithAudioData(float[] audioData, int numSamples, ReadableMap options) throws Exception {
         if (vadContext == 0) {
             throw new Exception("VAD context is null");
         }
 
-        // Decode base64 audio data to float array
-        byte[] audioBytes = Base64.decode(audioDataBase64, Base64.DEFAULT);
-        int numSamples = audioBytes.length / 4; // 4 bytes per float
-        float[] audioData = new float[numSamples];
-
-        for (int i = 0; i < numSamples; i++) {
-            int intBits = (audioBytes[i * 4] & 0xFF) |
-                         ((audioBytes[i * 4 + 1] & 0xFF) << 8) |
-                         ((audioBytes[i * 4 + 2] & 0xFF) << 16) |
-                         ((audioBytes[i * 4 + 3] & 0xFF) << 24);
-            audioData[i] = Float.intBitsToFloat(intBits);
-        }
-
         return processVadDetection(audioData, numSamples, options);
     }
 
-    public WritableArray detectSpeechFile(String filePathOrBase64, ReadableMap options) throws Exception {
-        if (vadContext == 0) {
-            throw new Exception("VAD context is null");
-        }
-
-        // Follow the same pattern as transcribeFile
-        String filePath = filePathOrBase64;
-
-        // Handle HTTP downloads
-        if (filePathOrBase64.startsWith("http://") || filePathOrBase64.startsWith("https://")) {
-            // Note: This would require access to the downloader, but for now we'll throw an error
-            throw new Exception("HTTP URLs not supported in VAD file detection. Please download the file first.");
-        }
-
-        float[] audioData;
-
-        // Check for resource identifier (bundled assets)
-        int resId = getResourceIdentifier(filePath);
-        if (resId > 0) {
-            audioData = AudioUtils.decodeWaveFile(reactContext.getResources().openRawResource(resId));
-        } else if (filePathOrBase64.startsWith("data:audio/wav;base64,")) {
-            // Handle base64 WAV data
-            audioData = AudioUtils.decodeWaveData(filePathOrBase64);
-        } else {
-            // Handle regular file path
-            audioData = AudioUtils.decodeWaveFile(new java.io.FileInputStream(new java.io.File(filePath)));
-        }
-
-        if (audioData == null) {
-            throw new Exception("Failed to load audio file: " + filePathOrBase64);
-        }
-
-        return processVadDetection(audioData, audioData.length, options);
-    }
-
-    public WritableArray detectSpeechWithAudioData(float[] audioData, ReadableMap options) throws Exception {
-        if (vadContext == 0) {
-            throw new Exception("VAD context is null");
-        }
-
-        return processVadDetection(audioData, audioData.length, options);
-    }
-
     private int getResourceIdentifier(String filePath) {
         int identifier = reactContext.getResources().getIdentifier(
             filePath,

package/ios/RNWhisper.mm

@@ -507,13 +507,16 @@ RCT_REMAP_METHOD(vadDetectSpeech,
     }
 
     // Decode base64 audio data
-    NSData *audioData = [[NSData alloc] initWithBase64EncodedString:audioDataBase64 options:0];
-    if (audioData == nil) {
+    NSData *pcmData = [[NSData alloc] initWithBase64EncodedString:audioDataBase64 options:0];
+    if (pcmData == nil) {
         reject(@"whisper_vad_error", @"Invalid audio data", nil);
         return;
     }
 
-    NSArray *segments = [vadContext detectSpeech:audioData options:options];
+    int count = 0;
+    float *data = [RNWhisperAudioUtils decodeWaveData:pcmData count:&count cutHeader:NO];
+
+    NSArray *segments = [vadContext detectSpeech:data samplesCount:count options:options];
     resolve(segments);
 }
 
@@ -549,10 +552,7 @@ RCT_REMAP_METHOD(vadDetectSpeechFile,
         return;
     }
 
-    // Convert float32 data to NSData for VAD context
-    NSData *audioData = [NSData dataWithBytes:data length:count * sizeof(float)];
-
-    NSArray *segments = [vadContext detectSpeech:audioData options:options];
+    NSArray *segments = [vadContext detectSpeech:data samplesCount:count options:options];
     resolve(segments);
 }
 

package/ios/RNWhisperVadContext.h

@@ -23,7 +23,7 @@
 - (NSString *)reasonNoMetal;
 - (struct whisper_vad_context *)getVadContext;
 - (dispatch_queue_t)getDispatchQueue;
-- (NSArray *)detectSpeech:(NSData *)audioData options:(NSDictionary *)options;
+- (NSArray *)detectSpeech:(float *)samples samplesCount:(int)samplesCount options:(NSDictionary *)options;
 - (void)invalidate;
 
 @end

package/ios/RNWhisperVadContext.mm

@@ -73,18 +73,14 @@
     return dQueue;
 }
 
-- (NSArray *)detectSpeech:(NSData *)audioData options:(NSDictionary *)options {
+- (NSArray *)detectSpeech:(float *)samples samplesCount:(int)samplesCount options:(NSDictionary *)options {
     if (vctx == NULL) {
         NSLog(@"VAD context is null");
         return @[];
     }
 
-    // Convert NSData to float array
-    const float *samples = (const float *)[audioData bytes];
-    int n_samples = (int)[audioData length] / sizeof(float);
-
     // Run VAD detection
-    bool speechDetected = whisper_vad_detect_speech(vctx, samples, n_samples);
+    bool speechDetected = whisper_vad_detect_speech(vctx, samples, samplesCount);
     if (!speechDetected) {
         return @[];
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "whisper.rn",
-  "version": "0.4.1",
+  "version": "0.4.2",
   "description": "React Native binding of whisper.cpp",
   "main": "lib/commonjs/index",
   "module": "lib/module/index",