feat(ios): add option to saving recorded audio as wav on startRealtim…

…eTranscribe
mybigday · Sep 23, 2023 · 750ab66 · 750ab66
1 parent 61f01e7
commit 750ab66
Show file tree

Hide file tree

Showing 5 changed files with 105 additions and 4 deletions.
diff --git a/ios/RNWhisper.xcodeproj/project.pbxproj b/ios/RNWhisper.xcodeproj/project.pbxproj
@@ -8,6 +8,9 @@
 
 /* Begin PBXBuildFile section */
 		5E555C0D2413F4C50049A1A2 /* RNWhisper.mm in Sources */ = {isa = PBXBuildFile; fileRef = B3E7B5891CC2AC0600A0062D /* RNWhisper.mm */; };
+		7FE0BBA12ABE6C7B0049B4E4 /* RNWhisperDownloader.m in Sources */ = {isa = PBXBuildFile; fileRef = 7FE0BB9B2ABE6C7B0049B4E4 /* RNWhisperDownloader.m */; };
+		7FE0BBA22ABE6C7B0049B4E4 /* RNWhisperAudioUtils.m in Sources */ = {isa = PBXBuildFile; fileRef = 7FE0BB9C2ABE6C7B0049B4E4 /* RNWhisperAudioUtils.m */; };
+		7FE0BBA32ABE6C7B0049B4E4 /* RNWhisperContext.mm in Sources */ = {isa = PBXBuildFile; fileRef = 7FE0BBA02ABE6C7B0049B4E4 /* RNWhisperContext.mm */; };
 /* End PBXBuildFile section */
 
 /* Begin PBXCopyFilesBuildPhase section */
@@ -24,6 +27,13 @@
 
 /* Begin PBXFileReference section */
 		134814201AA4EA6300B7C361 /* libRNWhisper.a */ = {isa = PBXFileReference; explicitFileType = archive.ar; includeInIndex = 0; path = libRNWhisper.a; sourceTree = BUILT_PRODUCTS_DIR; };
+		7FE0BB9A2ABE6C7B0049B4E4 /* RNWhisper.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = RNWhisper.h; sourceTree = "<group>"; };
+		7FE0BB9B2ABE6C7B0049B4E4 /* RNWhisperDownloader.m */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.objc; path = RNWhisperDownloader.m; sourceTree = "<group>"; };
+		7FE0BB9C2ABE6C7B0049B4E4 /* RNWhisperAudioUtils.m */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.objc; path = RNWhisperAudioUtils.m; sourceTree = "<group>"; };
+		7FE0BB9D2ABE6C7B0049B4E4 /* RNWhisperContext.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = RNWhisperContext.h; sourceTree = "<group>"; };
+		7FE0BB9E2ABE6C7B0049B4E4 /* RNWhisperDownloader.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = RNWhisperDownloader.h; sourceTree = "<group>"; };
+		7FE0BB9F2ABE6C7B0049B4E4 /* RNWhisperAudioUtils.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = RNWhisperAudioUtils.h; sourceTree = "<group>"; };
+		7FE0BBA02ABE6C7B0049B4E4 /* RNWhisperContext.mm */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.objcpp; path = RNWhisperContext.mm; sourceTree = "<group>"; };
 		B3E7B5891CC2AC0600A0062D /* RNWhisper.mm */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.objcpp; path = RNWhisper.mm; sourceTree = "<group>"; };
 /* End PBXFileReference section */
 
@@ -49,6 +59,13 @@
 		58B511D21A9E6C8500147676 = {
 			isa = PBXGroup;
 			children = (
+				7FE0BB9F2ABE6C7B0049B4E4 /* RNWhisperAudioUtils.h */,
+				7FE0BB9C2ABE6C7B0049B4E4 /* RNWhisperAudioUtils.m */,
+				7FE0BB9A2ABE6C7B0049B4E4 /* RNWhisper.h */,
+				7FE0BB9D2ABE6C7B0049B4E4 /* RNWhisperContext.h */,
+				7FE0BBA02ABE6C7B0049B4E4 /* RNWhisperContext.mm */,
+				7FE0BB9E2ABE6C7B0049B4E4 /* RNWhisperDownloader.h */,
+				7FE0BB9B2ABE6C7B0049B4E4 /* RNWhisperDownloader.m */,
 				B3E7B5891CC2AC0600A0062D /* RNWhisper.mm */,
 				134814211AA4EA7D00B7C361 /* Products */,
 			);
@@ -112,6 +129,9 @@
 			buildActionMask = 2147483647;
 			files = (
 				5E555C0D2413F4C50049A1A2 /* RNWhisper.mm in Sources */,
+				7FE0BBA22ABE6C7B0049B4E4 /* RNWhisperAudioUtils.m in Sources */,
+				7FE0BBA32ABE6C7B0049B4E4 /* RNWhisperContext.mm in Sources */,
+				7FE0BBA12ABE6C7B0049B4E4 /* RNWhisperDownloader.m in Sources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
@@ -223,9 +243,7 @@
 					"$(SRCROOT)/../../react-native/React/**",
 				);
 				LIBRARY_SEARCH_PATHS = "$(inherited)";
-				OTHER_LDFLAGS = (
-					"-ObjC",
-				);
+				OTHER_LDFLAGS = "-ObjC";
 				PRODUCT_NAME = RNWhisper;
 				SKIP_INSTALL = YES;
 			};

diff --git a/ios/RNWhisperAudioUtils.h b/ios/RNWhisperAudioUtils.h
@@ -0,0 +1,8 @@
+#import <Foundation/Foundation.h>
+
+@interface RNWhisperAudioUtils : NSObject
+
++ (NSData *)concatShortBuffers:(NSMutableArray<NSValue *> *)buffers sliceSize:(int)sliceSize lastSliceSize:(int)lastSliceSize;
++ (void)saveWavFile:(NSData *)rawData audioOutputFile:(NSString *)audioOutputFile;
+
+@end
diff --git a/ios/RNWhisperAudioUtils.m b/ios/RNWhisperAudioUtils.m
@@ -0,0 +1,64 @@
+#import "RNWhisperAudioUtils.h"
+#import "whisper.h"
+
+@implementation RNWhisperAudioUtils
+
++ (NSData *)concatShortBuffers:(NSMutableArray<NSValue *> *)buffers sliceSize:(int)sliceSize lastSliceSize:(int)lastSliceSize {
+    NSMutableData *outputData = [NSMutableData data];
+    for (NSValue *buffer in buffers) {
+        int size = sliceSize;
+        if (buffer == buffers.lastObject) {
+            size = lastSliceSize;
+        }
+        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];
+}
+
+@end
diff --git a/ios/RNWhisperContext.mm b/ios/RNWhisperContext.mm
@@ -1,4 +1,5 @@
 #import "RNWhisperContext.h"
+#import "RNWhisperAudioUtils.h"
 
 #define NUM_BYTES_PER_BUFFER 16 * 1024
 
@@ -212,6 +213,17 @@ - (void)fullTranscribeSamples:(RNWhisperContextRecordState*) state {
         NSLog(@"[RNWhisper] Transcribe end");
         result[@"isStoppedByAction"] = @(state->isStoppedByAction);
         result[@"isCapturing"] = @(false);
+
+        // Save wav if needed
+        if (state->options[@"audioOutputPath"] != nil) {
+            [RNWhisperAudioUtils
+                saveWavFile:[RNWhisperAudioUtils concatShortBuffers:state->shortBufferSlices
+                            sliceSize:state->audioSliceSec * WHISPER_SAMPLE_RATE
+                            lastSliceSize:nSamplesOfIndex]
+                audioOutputFile:state->options[@"audioOutputPath"]
+            ];
+        }
+
         state->transcribeHandler(state->jobId, @"end", result);
     } else if (code == 0) {
         result[@"isCapturing"] = @(true);

diff --git a/src/index.ts b/src/index.ts
@@ -60,7 +60,6 @@ export type TranscribeRealtimeOptions = TranscribeOptions & {
   realtimeAudioSliceSec?: number
   /**
    * Output path for audio file. If not set, the audio file will not be saved
-   * TODO: Support iOS
    * (Default: Undefined)
    */
   audioOutputPath?: string