react-native-davoice-tts 1.0.409 → 1.0.411

package/README.md

@@ -383,5 +383,25 @@ await speech.getVoiceAbilities("Rich");
 await speech.changeVoice("Rich");
 await speech.speak("Ok, [drill sergeant] Private, give me 50 push-ups! [regular] And you, Daniel, come with me.");
 ```
+#### Voice cloning (your own voices)
+
+Turn a recording into a voice: `createVoiceClone()` takes a wav, mp3, m4a,
+aac, ogg or flac file, keeps the best ~6 s of speech (silence trimmed, long
+pauses shortened, cut between words) and saves a scrambled `.emd` voice file.
+It needs the voice encoder for the model's language, `embedder_<code>.dm`
+(`embedder_en.dm` for English). Point `setExternalVoicesPath()` at the folder
+and the voice is listed and selected like a bundled one (the file name is the
+voice name; a bundled voice wins over a file with the same name).
+
+```ts
+const folder = `${RNFS.DocumentDirectoryPath}/voices`;
+await speech.setExternalVoicesPath(folder);
+const r = await speech.createVoiceClone(recordingPath, `${folder}/Me.emd`,
+  require('./assets/models/embedder_en.dm'));
+// r.warnings is non-empty when less than 4 s of speech remained.
+await speech.changeVoice('Me');
+await speech.speak('Hello in my own voice.');
+```
+
 The existing speak speed argument still applies, with pitch preserved.
 Rebuild the native app after updating the bundled AAR/XCFramework.

package/TTSRNBridge.podspec

@@ -2,7 +2,7 @@ require 'json'
 
 Pod::Spec.new do |s|
     s.name         = "TTSRNBridge"
-    s.version      = "1.0.405" # Update to your package version
+    s.version      = "1.0.409" # Update to your package version
     s.summary      = "TTS for React Native."
     s.description  = <<-DESC
                      A React Native module for tts .

package/android/libs/com/davoice/tts/1.0.0/tts-1.0.0.aar.md5

@@ -1 +1 @@
-135c22d998ff29b3e1aeabc5a1fd1392  tts-1.0.0.aar
+a619b1a82c872eb95700500b006c8456  tts-1.0.0.aar

package/android/libs/com/davoice/tts/1.0.0/tts-1.0.0.aar.sha1

@@ -1 +1 @@
-15e71c2af9e1a87f8a181497d0caafc0ef18e717  tts-1.0.0.aar
+b8d12c2f14c766d8f75e4cd81759a31e14848f54  tts-1.0.0.aar

package/android/src/main/java/com/davoice/tts/rn/DaVoiceTTSBridge.java

@@ -387,6 +387,73 @@ public class DaVoiceTTSBridge extends ReactContextBaseJavaModule {
         promise.resolve(abilities);
     }
 
+    private static String stripFileScheme(String path) {
+        return path != null && path.startsWith("file://") ? Uri.parse(path).getPath() : path;
+    }
+
+    // Any path form the app may pass (file path, file://, asset:/, http(s), RN release
+    // res/raw names, asset pack) -> readable local file, or null.
+    @Nullable
+    private File resolveAnyFile(String pathOrURL, @Nullable String defaultExt) {
+        if (pathOrURL == null || pathOrURL.trim().isEmpty()) return null;
+        File direct = new File(stripFileScheme(pathOrURL.trim()));
+        if (direct.isFile()) return direct;
+        try {
+            File f = resolveToLocalFileForRead(pathOrURL, defaultExt);
+            if (f != null && f.isFile()) return f;
+        } catch (Exception ignored) { }
+        File plain = resolvePlainModelToFile(pathOrURL.trim(), defaultExt);
+        return plain != null && plain.isFile() ? plain : null;
+    }
+
+    // Also look for voices (<name>.emd) in the folder `path`; null/"" removes them.
+    @ReactMethod
+    public void setExternalVoicesPath(@Nullable String path, Promise promise) {
+        WritableArray names = Arguments.createArray();
+        try {
+            if (tts != null) {
+                for (String name : tts.setExternalVoicesPath(stripFileScheme(path))) names.pushString(name);
+            }
+        } catch (Exception e) {
+            Log.w(TAG, "setExternalVoicesPath failed: " + e.getMessage());
+        }
+        promise.resolve(names);
+    }
+
+    // Reference audio (wav/mp3/m4a/...) -> scrambled .emd at outputPath, using the
+    // voice encoder embedder_<code>.dm. Runs off the bridge thread (takes seconds).
+    @ReactMethod
+    public void createVoiceClone(String inputAudio, String outputPath, String embedder, Promise promise) {
+        initExecutor.execute(() -> {
+            try {
+                if (inputAudio == null || outputPath == null || embedder == null) {
+                    promise.reject("invalid_args", "createVoiceClone(inputAudio, outputPath, embedder)");
+                    return;
+                }
+                String ext = inputAudio.contains(".") ? inputAudio.substring(inputAudio.lastIndexOf('.') + 1) : "wav";
+                File input = resolveAnyFile(inputAudio, ext.length() <= 5 ? ext : "wav");
+                if (input == null) { promise.reject("input_missing", "Input audio missing/unresolvable: " + inputAudio); return; }
+                File embedderFile = resolveAnyFile(embedder, "dm");
+                String embedderPath = embedderFile != null ? embedderFile.getAbsolutePath() : embedder;  // else an APK asset name
+                java.util.Map<String, Object> result =
+                    tts.createVoiceClone(input.getAbsolutePath(), stripFileScheme(outputPath), embedderPath);
+                WritableMap map = Arguments.createMap();
+                map.putString("path", String.valueOf(result.get("path")));
+                map.putString("voice", String.valueOf(result.get("voice")));
+                map.putDouble("seconds", ((Number) result.get("seconds")).doubleValue());
+                for (String key : new String[]{"warnings", "report"}) {
+                    WritableArray list = Arguments.createArray();
+                    Object value = result.get(key);
+                    if (value instanceof java.util.List) for (Object line : (java.util.List<?>) value) list.pushString(String.valueOf(line));
+                    map.putArray(key, list);
+                }
+                promise.resolve(map);
+            } catch (Exception e) {
+                promise.reject("voice_clone_failed", e.getMessage(), e);
+            }
+        });
+    }
+
     @ReactMethod
     public void changeVoice(String name, Promise promise) {
         try {

package/ios/SpeechBridge/SpeechBridge.m

@@ -1216,6 +1216,46 @@ RCT_EXPORT_METHOD(getVoiceAbilities:(NSString *)name
   resolve(self.tts && name ? ([self.tts getVoiceAbilities:name] ?: @[]) : @[]);
 }
 
+RCT_EXPORT_METHOD(setExternalVoicesPath:(NSString *)path
+                  resolver:(RCTPromiseResolveBlock)resolve
+                  rejecter:(RCTPromiseRejectBlock)reject)
+{
+  DaVoiceTTS *tts = self.tts;
+  if (!tts) { resolve(@[]); return; }
+  NSString *folder = (path && (id)path != [NSNull null] && path.length) ? path : nil;
+  if ([folder hasPrefix:@"file://"]) folder = [NSURL URLWithString:folder].path;
+  [tts setExternalVoicesPath:folder completion:^(NSArray<NSString *> *names) { resolve(names ?: @[]); }];
+}
+
+// Reference audio (wav/mp3/m4a/...) -> scrambled .emd at outputPath, using the
+// voice encoder embedder_<code>.dm. Runs off the bridge queue (takes seconds).
+RCT_EXPORT_METHOD(createVoiceClone:(NSString *)inputAudio
+                  outputPath:(NSString *)outputPath
+                  embedder:(NSString *)embedder
+                  resolver:(RCTPromiseResolveBlock)resolve
+                  rejecter:(RCTPromiseRejectBlock)reject)
+{
+  if (!inputAudio || !outputPath || !embedder || (id)outputPath == [NSNull null]) {
+    reject(@"invalid_args", @"createVoiceClone(inputAudio, outputPath, embedder)", nil);
+    return;
+  }
+  dispatch_async(dispatch_get_global_queue(QOS_CLASS_USER_INITIATED, 0), ^{
+    BOOL remoteInput = [inputAudio hasPrefix:@"http://"] || [inputAudio hasPrefix:@"https://"];
+    BOOL remoteEmbedder = [embedder hasPrefix:@"http://"] || [embedder hasPrefix:@"https://"];
+    NSURL *inputURL = [self resolveLocalURLFromPathOrURL:inputAudio];
+    NSURL *embedderURL = [self resolveLocalURLFromPathOrURL:embedder];
+    NSString *output = [outputPath hasPrefix:@"file://"] ? [NSURL URLWithString:outputPath].path : outputPath;
+    NSError *error = nil;
+    NSDictionary *result = (inputURL && embedderURL)
+      ? [DaVoiceTTS createVoiceClone:inputURL.path outputPath:output embedderPath:embedderURL.path error:&error]
+      : nil;
+    if (remoteInput && inputURL) [[NSFileManager defaultManager] removeItemAtURL:inputURL error:nil];
+    if (remoteEmbedder && embedderURL) [[NSFileManager defaultManager] removeItemAtURL:embedderURL error:nil];
+    if (result) resolve(result);
+    else reject(@"voice_clone_failed", error.localizedDescription ?: @"Could not resolve the input audio or embedder path", error);
+  });
+}
+
 RCT_EXPORT_METHOD(changeVoice:(NSString *)name
                   resolver:(RCTPromiseResolveBlock)resolve
                   rejecter:(RCTPromiseRejectBlock)reject)

package/ios/TTSRNBridge/DaVoiceTTSBridge.m

@@ -289,6 +289,46 @@ RCT_EXPORT_METHOD(getVoiceAbilities:(NSString *)name
   resolve(self.ttsInstance && name ? ([self.ttsInstance getVoiceAbilities:name] ?: @[]) : @[]);
 }
 
+RCT_EXPORT_METHOD(setExternalVoicesPath:(NSString *)path
+                  resolver:(RCTPromiseResolveBlock)resolve
+                  rejecter:(RCTPromiseRejectBlock)reject)
+{
+  DaVoiceTTS *tts = self.ttsInstance;
+  if (!tts) { resolve(@[]); return; }
+  NSString *folder = (path && (id)path != [NSNull null] && path.length) ? path : nil;
+  if ([folder hasPrefix:@"file://"]) folder = [NSURL URLWithString:folder].path;
+  [tts setExternalVoicesPath:folder completion:^(NSArray<NSString *> *names) { resolve(names ?: @[]); }];
+}
+
+// Reference audio (wav/mp3/m4a/...) -> scrambled .emd at outputPath, using the
+// voice encoder embedder_<code>.dm. Runs off the bridge queue (takes seconds).
+RCT_EXPORT_METHOD(createVoiceClone:(NSString *)inputAudio
+                  outputPath:(NSString *)outputPath
+                  embedder:(NSString *)embedder
+                  resolver:(RCTPromiseResolveBlock)resolve
+                  rejecter:(RCTPromiseRejectBlock)reject)
+{
+  if (!inputAudio || !outputPath || !embedder || (id)outputPath == [NSNull null]) {
+    reject(@"invalid_args", @"createVoiceClone(inputAudio, outputPath, embedder)", nil);
+    return;
+  }
+  dispatch_async(dispatch_get_global_queue(QOS_CLASS_USER_INITIATED, 0), ^{
+    BOOL remoteInput = [inputAudio hasPrefix:@"http://"] || [inputAudio hasPrefix:@"https://"];
+    BOOL remoteEmbedder = [embedder hasPrefix:@"http://"] || [embedder hasPrefix:@"https://"];
+    NSURL *inputURL = [self resolveLocalURLFromPathOrURL:inputAudio];
+    NSURL *embedderURL = [self resolveLocalURLFromPathOrURL:embedder];
+    NSString *output = [outputPath hasPrefix:@"file://"] ? [NSURL URLWithString:outputPath].path : outputPath;
+    NSError *error = nil;
+    NSDictionary *result = (inputURL && embedderURL)
+      ? [DaVoiceTTS createVoiceClone:inputURL.path outputPath:output embedderPath:embedderURL.path error:&error]
+      : nil;
+    if (remoteInput && inputURL) [[NSFileManager defaultManager] removeItemAtURL:inputURL error:nil];
+    if (remoteEmbedder && embedderURL) [[NSFileManager defaultManager] removeItemAtURL:embedderURL error:nil];
+    if (result) resolve(result);
+    else reject(@"voice_clone_failed", error.localizedDescription ?: @"Could not resolve the input audio or embedder path", error);
+  });
+}
+
   RCT_EXPORT_METHOD(changeVoice:(NSString *)name
                   resolver:(RCTPromiseResolveBlock)resolve
                   rejecter:(RCTPromiseRejectBlock)reject)

package/ios/TTSRNBridge/DavoiceTTS.xcframework/ios-arm64/DavoiceTTS.framework/Headers/DavoiceTTS-Swift.h

@@ -361,6 +361,36 @@ SWIFT_CLASS("_TtC10DavoiceTTS10DaVoiceTTS")
 /// 50! [regular] Thanks.” Unmatched tags use the default voice; tags are never
 /// spoken. Empty for voices without variants, v1 models, or after destroy().
 - (NSArray<NSString *> * _Nonnull)getVoiceAbilities:(NSString * _Nonnull)name SWIFT_WARN_UNUSED_RESULT;
+/// Also look for voices in the folder <code>path</code>: <name>.emd / <name>.npy files,
+/// e.g. made by createVoiceClone(). They are listed by getAllVoices()
+/// after the bundled voices and selected with changeVoice() like them; a
+/// bundled voice wins over an external file with the same name. Call it
+/// again to pick up new files; nil or “” removes the external voices.
+/// <code>completion</code> gets the external voice names found ([] for a v1 model or a
+/// missing folder).
+- (void)setExternalVoicesPath:(NSString * _Nullable)path completion:(void (^ _Nullable)(NSArray<NSString *> * _Nonnull))completion;
+/// Creates a voice clone from reference audio and saves it as a
+/// scrambled .emd at <code>outputPath</code>.
+/// <ul>
+///   <li>
+///     inputAudioPath: wav, mp3, m4a/aac, caf, aiff, flac, … The best ~6 s
+///     of speech is used: silence trimmed, long pauses shortened, cut at a gap
+///     between words (never inside a word).
+///   </li>
+///   <li>
+///     outputPath: the .emd file to write (”.emd” is added when missing), or a
+///     folder (-> <folder>/<input name>.emd). The file name is the voice name.
+///   </li>
+///   <li>
+///     embedderPath: the language’s voice encoder, embedder_<code>.dm
+///     (e.g. embedder_en.dm); it is unscrambled in memory only.
+///   </li>
+/// </ul>
+/// Returns [“path”, “voice”, “seconds”, “warnings”, “report”]; “warnings” is
+/// non-empty when less than 4 s of speech remained or the cut had to fall
+/// inside speech. Put the file in the folder given to setExternalVoicesPath()
+/// (and call that again) to use it.
++ (NSDictionary<NSString *, id> * _Nullable)createVoiceClone:(NSString * _Nonnull)inputAudioPath outputPath:(NSString * _Nonnull)outputPath embedderPath:(NSString * _Nonnull)embedderPath error:(NSError * _Nullable * _Nullable)error SWIFT_WARN_UNUSED_RESULT;
 - (void)changeVoice:(NSString * _Nonnull)name completion:(void (^ _Nonnull)(NSInteger))completion;
 @property (nonatomic, copy) void (^ _Nullable onLastUtteranceFinished)(void);
 + (BOOL)activateLicenseWithLicenseKey:(NSString * _Nonnull)licenseKey SWIFT_WARN_UNUSED_RESULT;