Add clamping speed parameter

Author: IgorSwat

packages/react-native-executorch/common/rnexecutorch/models/text_to_speech/kokoro/Constants.h

@@ -20,6 +20,8 @@ inline constexpr size_t kMinDurationTicks =
 inline constexpr size_t kMaxDurationTicks =
     296; // Corresponds to DurationPredictor output and one of Synthesizer's
          // input shapes
+inline constexpr float kMinValidSpeed = 0.1F;
+inline constexpr float kMaxValidSpeed = 3.0F;
 
 // Model input sizes - voice reference vector
 inline constexpr int32_t kVoiceRefSize =

packages/react-native-executorch/common/rnexecutorch/models/text_to_speech/kokoro/Kokoro.cpp

@@ -96,6 +96,20 @@ std::vector<float> Kokoro::generate(std::u32string input, float speed,
                             "Kokoro: maximum input text size exceeded");
   }
 
+  if (speed < constants::kMinValidSpeed) {
+    throw RnExecutorchError(RnExecutorchErrorCode::InvalidUserInput,
+                            "Kokoro: speed value too low (min " +
+                                std::to_string(constants::kMinValidSpeed) +
+                                ")");
+  }
+
+  if (speed > constants::kMaxValidSpeed) {
+    throw RnExecutorchError(RnExecutorchErrorCode::InvalidUserInput,
+                            "Kokoro: speed value too high (max " +
+                                std::to_string(constants::kMaxValidSpeed) +
+                                ")");
+  }
+
   if (input.empty()) {
     return {};
   }
@@ -133,6 +147,20 @@ std::vector<float> Kokoro::generate(std::u32string input, float speed,
 
 void Kokoro::stream(std::shared_ptr<jsi::Function> callback, float speed,
                     bool phonemize, bool stopOnEmptyBuffer) {
+  if (speed < constants::kMinValidSpeed) {
+    throw RnExecutorchError(RnExecutorchErrorCode::InvalidUserInput,
+                            "Kokoro: speed value too low (min " +
+                                std::to_string(constants::kMinValidSpeed) +
+                                ")");
+  }
+
+  if (speed > constants::kMaxValidSpeed) {
+    throw RnExecutorchError(RnExecutorchErrorCode::InvalidUserInput,
+                            "Kokoro: speed value too high (max " +
+                                std::to_string(constants::kMaxValidSpeed) +
+                                ")");
+  }
+
   // Create a callback
   auto nativeCallback = [this, callback](const std::vector<float> &audioVec) {
     if (this->isStreaming_) {

packages/react-native-executorch/src/modules/natural_language_processing/TextToSpeechModule.ts

@@ -75,21 +75,28 @@ export class TextToSpeechModule {
       ...sources
     );
 
-    if (paths === null || paths.length !== sources.length) {
+    // Required fields
+    const [duration, synth, voice] = paths;
+    if (!duration || !synth || !voice) {
       throw new RnExecutorchError(
         RnExecutorchErrorCode.DownloadInterrupted,
-        'Download interrupted or missing resource.'
+        'Kokoro: missing required model paths.'
       );
     }
 
+    // Optional fields
+    const tagger = paths[taggerIdx] ?? '';
+    const lexicon = paths[lexiconIdx] ?? '';
+    const neural = paths[neuralModelIdx] ?? '';
+
     return await global.loadTextToSpeechKokoro(
       phonemizerConfig.lang,
-      taggerIdx >= 0 ? (paths[taggerIdx] as string) : '',
-      lexiconIdx >= 0 ? (paths[lexiconIdx] as string) : '',
-      neuralModelIdx >= 0 ? (paths[neuralModelIdx] as string) : '',
-      paths[0] as string, // DurationPredictor source
-      paths[1] as string, // Synthesizer source
-      paths[2] as string // Voice source
+      tagger,
+      lexicon,
+      neural,
+      duration,
+      synth,
+      voice
     );
   }