whisper.rn 0.4.0-rc.1 → 0.4.0-rc.11

package/ios/RNWhisperAudioUtils.m

@@ -3,60 +3,23 @@
 
 @implementation RNWhisperAudioUtils
 
-+ (NSData *)concatShortBuffers:(NSMutableArray<NSValue *> *)buffers sliceNSamples:(NSMutableArray<NSNumber *> *)sliceNSamples {
-    NSMutableData *outputData = [NSMutableData data];
-    for (int i = 0; i < buffers.count; i++) {
-        int size = [sliceNSamples objectAtIndex:i].intValue;
-        NSValue *buffer = [buffers objectAtIndex:i];
-        short *bufferPtr = buffer.pointerValue;
-        [outputData appendBytes:bufferPtr length:size * sizeof(short)];
-    }
-    return outputData;
-}
-
-+ (void)saveWavFile:(NSData *)rawData audioOutputFile:(NSString *)audioOutputFile {
-    NSMutableData *outputData = [NSMutableData data];
-    
-    // WAVE header
-    [outputData appendData:[@"RIFF" dataUsingEncoding:NSUTF8StringEncoding]]; // chunk id
-    int chunkSize = CFSwapInt32HostToLittle(36 + rawData.length);
-    [outputData appendBytes:&chunkSize length:sizeof(chunkSize)];
-    [outputData appendData:[@"WAVE" dataUsingEncoding:NSUTF8StringEncoding]]; // format
-    [outputData appendData:[@"fmt " dataUsingEncoding:NSUTF8StringEncoding]]; // subchunk 1 id
-    
-    int subchunk1Size = CFSwapInt32HostToLittle(16);
-    [outputData appendBytes:&subchunk1Size length:sizeof(subchunk1Size)];
-
-    short audioFormat = CFSwapInt16HostToLittle(1); // PCM
-    [outputData appendBytes:&audioFormat length:sizeof(audioFormat)];
-
-    short numChannels = CFSwapInt16HostToLittle(1); // mono
-    [outputData appendBytes:&numChannels length:sizeof(numChannels)];
-
-    int sampleRate = CFSwapInt32HostToLittle(WHISPER_SAMPLE_RATE);
-    [outputData appendBytes:&sampleRate length:sizeof(sampleRate)];
-
-    // (bitDepth * sampleRate * channels) >> 3
-    int byteRate = CFSwapInt32HostToLittle(WHISPER_SAMPLE_RATE * 1 * 16 / 8);
-    [outputData appendBytes:&byteRate length:sizeof(byteRate)];
-
-    // (bitDepth * channels) >> 3
-    short blockAlign = CFSwapInt16HostToLittle(16 / 8);
-    [outputData appendBytes:&blockAlign length:sizeof(blockAlign)];
-
-    // bitDepth
-    short bitsPerSample = CFSwapInt16HostToLittle(16);
-    [outputData appendBytes:&bitsPerSample length:sizeof(bitsPerSample)];
-
-    [outputData appendData:[@"data" dataUsingEncoding:NSUTF8StringEncoding]]; // subchunk 2 id
-    int subchunk2Size = CFSwapInt32HostToLittle((int)rawData.length);
-    [outputData appendBytes:&subchunk2Size length:sizeof(subchunk2Size)];
-
-    // Audio data
-    [outputData appendData:rawData];
-    
-    // Save to file
-    [outputData writeToFile:audioOutputFile atomically:YES];
++ (float *)decodeWaveData:(NSData*)data count:(int *)count cutHeader:(BOOL)cutHeader {
+  NSData *waveData = data;
+  if (cutHeader) {
+    // just cut 44 bytes from the beginning
+    waveData = [data subdataWithRange:NSMakeRange(44, [data length]-44)];
+  }
+  const short *shortArray = (const short *)[waveData bytes];
+  int shortCount = (int) ([waveData length] / sizeof(short));
+  float *floatArray = (float *) malloc(shortCount * sizeof(float));
+  for (NSInteger i = 0; i < shortCount; i++) {
+      float floatValue = ((float)shortArray[i]) / 32767.0;
+      floatValue = MAX(floatValue, -1.0);
+      floatValue = MIN(floatValue, 1.0);
+      floatArray[i] = floatValue;
+  }
+  *count = shortCount;
+  return floatArray;
 }
 
 + (float *)decodeWaveFile:(NSString*)filePath count:(int *)count {
@@ -65,19 +28,7 @@
     if (fileData == nil) {
         return nil;
     }
-    NSMutableData *waveData = [[NSMutableData alloc] init];
-    [waveData appendData:[fileData subdataWithRange:NSMakeRange(44, [fileData length]-44)]];
-    const short *shortArray = (const short *)[waveData bytes];
-    int shortCount = (int) ([waveData length] / sizeof(short));
-    float *floatArray = (float *) malloc(shortCount * sizeof(float));
-    for (NSInteger i = 0; i < shortCount; i++) {
-        float floatValue = ((float)shortArray[i]) / 32767.0;
-        floatValue = MAX(floatValue, -1.0);
-        floatValue = MIN(floatValue, 1.0);
-        floatArray[i] = floatValue;
-    }
-    *count = shortCount;
-    return floatArray;
+    return [RNWhisperAudioUtils decodeWaveData:fileData count:count cutHeader:YES];
 }
 
 @end

package/ios/RNWhisperContext.h

@@ -11,22 +11,20 @@
 
 typedef struct {
     __unsafe_unretained id mSelf;
-
-    int jobId;
     NSDictionary* options;
 
+    struct rnwhisper::job * job;
+
     bool isTranscribing;
     bool isRealtime;
     bool isCapturing;
     bool isStoppedByAction;
-    int maxAudioSec;
     int nSamplesTranscribing;
-    NSMutableArray<NSValue *> *shortBufferSlices;
-    NSMutableArray<NSNumber *> *sliceNSamples;
+    std::vector<int> sliceNSamples;
     bool isUseSlices;
     int sliceIndex;
     int transcribeSliceIndex;
-    int audioSliceSec;
+    NSString* audioOutputPath;
 
     AudioQueueRef queue;
     AudioStreamBasicDescription dataFormat;
@@ -40,15 +38,19 @@ typedef struct {
     dispatch_queue_t dQueue;
     struct whisper_context * ctx;
     RNWhisperContextRecordState recordState;
+    NSString * reasonNoMetal;
+    bool isMetalEnabled;
 }
 
-+ (instancetype)initWithModelPath:(NSString *)modelPath contextId:(int)contextId noCoreML:(BOOL)noCoreML;
++ (instancetype)initWithModelPath:(NSString *)modelPath contextId:(int)contextId noCoreML:(BOOL)noCoreML noMetal:(BOOL)noMetal useFlashAttn:(BOOL)useFlashAttn;
+- (bool)isMetalEnabled;
+- (NSString *)reasonNoMetal;
 - (struct whisper_context *)getContext;
 - (dispatch_queue_t)getDispatchQueue;
 - (OSStatus)transcribeRealtime:(int)jobId
     options:(NSDictionary *)options
     onTranscribe:(void (^)(int, NSString *, NSDictionary *))onTranscribe;
-- (void)transcribeFile:(int)jobId
+- (void)transcribeData:(int)jobId
     audioData:(float *)audioData
     audioDataCount:(int)audioDataCount
     options:(NSDictionary *)options
@@ -61,6 +63,7 @@ typedef struct {
 - (bool)isTranscribing;
 - (bool)isStoppedByAction;
 - (NSMutableDictionary *)getTextSegments;
+- (NSString *)bench:(int)maxThreads;
 - (void)invalidate;
 
 @end