react-native-nitro-onnx 0.1.0

package/cpp/VadEngine.hpp

@@ -0,0 +1,126 @@
+// ------------------------------------------------------------------------------
+// VadEngine.hpp
+// Voice Activity Detection wrapper with a sliding pre-buffer queue.
+//
+// Problem: sherpa-onnx emits onSpeechStart asynchronously after it has already
+// processed the first few frames of speech. The leading audio is therefore
+// missing from the segment returned at onSpeechEnd.
+//
+// Solution: keep a fixed-size sliding buffer of raw PCM. When speech starts,
+// the pre-buffer is prepended to the VAD output, preserving the first frames.
+// ------------------------------------------------------------------------------
+#pragma once
+
+#include "AudioUtils.hpp"
+#include "ThreadPool.hpp"
+
+#include <condition_variable>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <queue>
+#include <thread>
+#include <vector>
+
+// Forward declaration for sherpa-onnx C API.
+struct SherpaOnnxVoiceActivityDetector;
+
+namespace margelo::nitro::onnx::speech {
+
+/** Native VAD configuration before conversion from the generated VadConfig. */
+struct VadEngineConfig {
+  std::string modelPath;
+  float threshold = 0.5f;
+  float minSilenceDuration = 0.5f;
+  float minSpeechDuration = 0.25f;
+  int32_t preBufferMs = 300;
+};
+
+/** One buffered or emitted speech segment with raw float samples. */
+struct VadEngineSegment {
+  float startMs = 0.0f;
+  float endMs = 0.0f;
+  std::vector<float> samples;
+};
+
+/** Consumer callbacks. Implemented by the Nitro hybrid object. */
+class VadListener {
+ public:
+  virtual ~VadListener() = default;
+  virtual void onSpeechStart(const VadEngineSegment& segment) = 0;
+  virtual void onSpeechEnd(const VadEngineSegment& segment) = 0;
+  virtual void onError(const std::string& error) = 0;
+};
+
+/**
+ * Thread-safe VAD engine.
+ *
+ * All heavy work (sherpa-onnx inference) is performed on a background thread
+ * so that JS never blocks. A dedicated processor thread consumes audio from
+ * an input queue and emits segments through the listener.
+ */
+class VadEngine final {
+ public:
+  explicit VadEngine(std::shared_ptr<ThreadPool> threadPool);
+  ~VadEngine();
+
+  VadEngine(const VadEngine&) = delete;
+  VadEngine& operator=(const VadEngine&) = delete;
+
+  /** Initialize the sherpa-onnx VAD. Must be called before processing. */
+  void initialize(const VadEngineConfig& config, std::shared_ptr<VadListener> listener);
+
+  /** Return true if the VAD has been initialized. */
+  bool isInitialized() const;
+
+  /** Feed a chunk of 16 kHz mono f32 PCM audio. Non-blocking. */
+  void acceptWaveform(const std::vector<float>& samples);
+
+  /** Return any buffered segments without waiting for speech end. */
+  std::vector<VadEngineSegment> pullSegments();
+
+  /** Reset VAD state and pre-buffer. */
+  void reset();
+
+  /** Release native resources. */
+  void dispose();
+
+ private:
+  void processLoop();
+  void flushPreBuffer(std::vector<float>& target);
+  void emitSegment();
+
+  std::shared_ptr<ThreadPool> threadPool_;
+  std::shared_ptr<VadListener> listener_;
+
+  const SherpaOnnxVoiceActivityDetector* vad_ = nullptr;
+  VadEngineConfig config_;
+  std::atomic<bool> initialized_{false};
+  std::atomic<bool> stop_{false};
+  std::atomic<bool> resetRequested_{false};
+
+  // Input queue.
+  std::mutex inputMutex_;
+  std::condition_variable inputCv_;
+  std::queue<std::vector<float>> inputQueue_;
+
+  // Sliding pre-buffer. Always holds the most recent N milliseconds of audio.
+  std::mutex preBufferMutex_;
+  std::vector<float> preBuffer_;
+  size_t preBufferCapacity_ = 0;
+
+  // Output segments waiting to be pulled.
+  std::mutex outputMutex_;
+  std::vector<VadEngineSegment> pendingSegments_;
+
+  // Processor thread.
+  std::thread processorThread_;
+
+  // Stream time tracking.
+  float streamMs_ = 0.0f;
+  bool inSpeech_ = false;
+  float currentSpeechStartMs_ = 0.0f;
+  std::vector<float> currentSpeechSamples_;
+};
+
+}  // namespace margelo::nitro::onnx::speech

package/ios/OnnxSpeechInitializer.mm

@@ -0,0 +1,37 @@
+// ------------------------------------------------------------------------------
+// OnnxSpeechInitializer.mm
+// ------------------------------------------------------------------------------
+// Sets up resource and cache directories for the C++ layer on iOS.
+// Runs at +load time so paths are available before any HybridObject is created.
+// ------------------------------------------------------------------------------
+#import <Foundation/Foundation.h>
+
+#include "ResourceDir.hpp"
+
+@interface NitroOnnxSpeechInitializer : NSObject
+@end
+
+@implementation NitroOnnxSpeechInitializer
+
++ (void)load {
+  NSBundle* mainBundle = [NSBundle mainBundle];
+
+  // The podspec bundles assets into NitroOnnxSpeech_Resources.bundle.
+  NSString* resourceBundlePath = [mainBundle pathForResource:@"NitroOnnxSpeech_Resources"
+                                                      ofType:@"bundle"];
+  if (resourceBundlePath) {
+    margelo::nitro::onnx::speech::setResourceDir([resourceBundlePath UTF8String]);
+  } else {
+    // Fallback: model may be copied directly into the main bundle.
+    margelo::nitro::onnx::speech::setResourceDir([[mainBundle resourcePath] UTF8String]);
+  }
+
+  NSArray<NSString*>* cachePaths = NSSearchPathForDirectoriesInDomains(
+      NSCachesDirectory, NSUserDomainMask, YES);
+  NSString* cacheDir = cachePaths.firstObject;
+  if (cacheDir) {
+    margelo::nitro::onnx::speech::setCacheDir([cacheDir UTF8String]);
+  }
+}
+
+@end

package/ios/PrivacyInfo.xcprivacy

@@ -0,0 +1,14 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+  <key>NSPrivacyTracking</key>
+  <false/>
+  <key>NSPrivacyTrackingDomains</key>
+  <array/>
+  <key>NSPrivacyCollectedDataTypes</key>
+  <array/>
+  <key>NSPrivacyAccessedAPITypes</key>
+  <array/>
+</dict>
+</plist>

package/lib/index.d.ts

@@ -0,0 +1,11 @@
+import type { AudioFormat, OnnxSpeech } from "./specs/OnnxSpeech.nitro.js";
+export type { AudioFormat, Result, VadConfig, VadSegment, VadEvents, Vad, AsrModelType, AsrModelConfig, AsrResult, StreamingAsrEvents, OfflineAsr, StreamingAsr, TtsModelType, TtsModelConfig, TtsResult, Tts, SpeakerEmbeddingConfig, RegisteredSpeaker, SpeakerManager, OnnxSpeech, } from "./specs/OnnxSpeech.nitro.js";
+/**
+ * Get the shared OnnxSpeech Nitro module instance.
+ * Heavy AI models are loaded on demand and cached as singletons behind
+ * the individual hybrid objects (Vad, Asr, Tts, SpeakerManager).
+ */
+export declare function getOnnxSpeech(): OnnxSpeech;
+/** Default 16 kHz mono f32 PCM format used by every audio interface. */
+export declare const DEFAULT_AUDIO_FORMAT: AudioFormat;
+//# sourceMappingURL=index.d.ts.map

package/lib/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,EAAE,WAAW,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE3E,YAAY,EACV,WAAW,EACX,MAAM,EACN,SAAS,EACT,UAAU,EACV,SAAS,EACT,GAAG,EACH,YAAY,EACZ,cAAc,EACd,SAAS,EACT,kBAAkB,EAClB,UAAU,EACV,YAAY,EACZ,YAAY,EACZ,cAAc,EACd,SAAS,EACT,GAAG,EACH,sBAAsB,EACtB,iBAAiB,EACjB,cAAc,EACd,UAAU,GACX,MAAM,6BAA6B,CAAC;AAErC;;;;GAIG;AACH,wBAAgB,aAAa,IAAI,UAAU,CAE1C;AAED,wEAAwE;AACxE,eAAO,MAAM,oBAAoB,EAAE,WAIlC,CAAC"}

package/lib/index.js

@@ -0,0 +1,19 @@
+// ------------------------------------------------------------------------------
+// Public entry point for react-native-nitro-onnx.
+// ------------------------------------------------------------------------------
+import { NitroModules } from "react-native-nitro-modules";
+/**
+ * Get the shared OnnxSpeech Nitro module instance.
+ * Heavy AI models are loaded on demand and cached as singletons behind
+ * the individual hybrid objects (Vad, Asr, Tts, SpeakerManager).
+ */
+export function getOnnxSpeech() {
+    return NitroModules.createHybridObject("NitroOnnxSpeech");
+}
+/** Default 16 kHz mono f32 PCM format used by every audio interface. */
+export const DEFAULT_AUDIO_FORMAT = {
+    sampleRate: 16000,
+    channels: 1,
+    sampleFormat: "f32le",
+};
+//# sourceMappingURL=index.js.map

package/lib/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,iFAAiF;AACjF,kDAAkD;AAClD,iFAAiF;AAEjF,OAAO,EAAE,YAAY,EAAE,MAAM,4BAA4B,CAAC;AA0B1D;;;;GAIG;AACH,MAAM,UAAU,aAAa;IAC3B,OAAO,YAAY,CAAC,kBAAkB,CAAa,iBAAiB,CAAC,CAAC;AACxE,CAAC;AAED,wEAAwE;AACxE,MAAM,CAAC,MAAM,oBAAoB,GAAgB;IAC/C,UAAU,EAAE,KAAK;IACjB,QAAQ,EAAE,CAAC;IACX,YAAY,EAAE,OAAO;CACtB,CAAC"}

package/lib/specs/OnnxSpeech.nitro.d.ts

@@ -0,0 +1,264 @@
+import type { HybridObject } from "react-native-nitro-modules";
+/** All speech hybrid objects are implemented in C++ on both platforms. */
+export type SpeechPlatforms = {
+    ios: "c++";
+    android: "c++";
+};
+/** Audio format constants used across the module. */
+export interface AudioFormat {
+    /** Sample rate in Hz. Always 16000 for this module. */
+    sampleRate: number;
+    /** Number of channels. Always 1 (mono). */
+    channels: number;
+    /** PCM sample format. */
+    sampleFormat: "f32le" | "s16le";
+}
+/** Generic result wrapper for operations that may fail on the native side. */
+export interface Result<T> {
+    success: boolean;
+    data?: T;
+    error?: string;
+}
+export interface VadConfig {
+    /**
+     * Optional path to a custom Silero VAD onnx model.
+     * If omitted, the bundled silero_vad.onnx resource is used.
+     */
+    modelPath?: string;
+    /** Threshold for speech start detection, range 0..1. Default: 0.5. */
+    threshold?: number;
+    /** Minimum silence duration in milliseconds before speech ends. Default: 500. */
+    minSilenceDurationMs?: number;
+    /** Minimum speech duration in milliseconds. Default: 250. */
+    minSpeechDurationMs?: number;
+    /** How many milliseconds of audio to keep before onSpeechStart. Default: 300. */
+    preBufferMs?: number;
+}
+/** VAD segment delivered after speech ends or on explicit pull. */
+export interface VadSegment {
+    /** Start offset in milliseconds relative to stream start. */
+    startMs: number;
+    /** End offset in milliseconds. */
+    endMs: number;
+    /** Audio samples as 16 kHz mono f32 PCM. */
+    samples: ArrayBuffer;
+}
+/** VAD events delivered asynchronously to JS. */
+export interface VadEvents {
+    onSpeechStart?: (segment: VadSegment) => void;
+    onSpeechEnd?: (segment: VadSegment) => void;
+    onError?: (error: string) => void;
+}
+/** Voice Activity Detection hybrid object. */
+export interface Vad extends HybridObject<SpeechPlatforms> {
+    /** Optional callback invoked when speech starts. */
+    onSpeechStart?: (segment: VadSegment) => void;
+    /** Optional callback invoked when speech ends. */
+    onSpeechEnd?: (segment: VadSegment) => void;
+    /** Optional callback invoked on VAD errors. */
+    onError?: (error: string) => void;
+    /** Initialize the VAD engine. Must be called before process. */
+    initialize(config: VadConfig): Promise<void>;
+    /** Returns true if the engine is ready to process audio. */
+    isInitialized(): boolean;
+    /** Feed a chunk of 16 kHz mono f32 PCM audio. Runs on a background thread. */
+    process(samples: ArrayBuffer): Promise<void>;
+    /** Pull any buffered speech segment(s) without waiting for speech end. */
+    pullSegments(): Promise<VadSegment[]>;
+    /** Reset internal state and buffers. */
+    reset(): Promise<void>;
+}
+export type AsrModelType = "whisper" | "transducer" | "paraformer" | "zipformer" | "conformer" | "wenet" | "telespeech" | "moonshine" | "dolphin" | "nemo" | "sense_voice";
+export interface AsrModelConfig {
+    type: AsrModelType;
+    /** Directory containing all model files. */
+    modelDir: string;
+    /** Path to tokens.txt or equivalent vocabulary file. */
+    tokensPath: string;
+    /** Whisper: encoder + decoder ONNX files. */
+    whisperEncoder?: string;
+    whisperDecoder?: string;
+    /** Transducer / Zipformer / Conformer: encoder/decoder/joiner. */
+    encoder?: string;
+    decoder?: string;
+    joiner?: string;
+    /** Paraformer / Wenet / Telespeech / SenseVoice: single model file. */
+    model?: string;
+    /** NeMo config file (.yaml). */
+    config?: string;
+    /** Number of threads for ONNX Runtime. Default: 2. */
+    numThreads?: number;
+    /** Decode method, e.g. "greedy_search" or "modified_beam_search". */
+    decodingMethod?: string;
+    /** Max active paths for beam search. */
+    maxActivePaths?: number;
+    /** Whisper language hint, e.g. "en", "zh". */
+    language?: string;
+    /** SenseVoice: whether to use itn. */
+    useItn?: boolean;
+}
+export interface AsrResult {
+    text: string;
+    /** Confidence score when available. */
+    score?: number;
+    /** Start offset in milliseconds. */
+    startMs?: number;
+    /** End offset in milliseconds. */
+    endMs?: number;
+    /** Word / token timestamps when supported by the model. */
+    timestamps?: number[];
+    /** Raw JSON metadata from the recognizer. */
+    json?: string;
+}
+export interface StreamingAsrEvents {
+    onPartialResult?: (result: AsrResult) => void;
+    onFinalResult?: (result: AsrResult) => void;
+    onError?: (error: string) => void;
+}
+/** Offline (file/chunk) ASR hybrid object. */
+export interface OfflineAsr extends HybridObject<SpeechPlatforms> {
+    /** Preload the model singleton. */
+    load(config: AsrModelConfig): Promise<void>;
+    isLoaded(): boolean;
+    /** Recognize a full buffer of 16 kHz mono f32 PCM. */
+    recognize(samples: ArrayBuffer): Promise<AsrResult>;
+    /** Recognize from a file path (native side reads the audio). */
+    recognizeFile(path: string): Promise<AsrResult>;
+    unload(): Promise<void>;
+}
+/** Streaming ASR hybrid object. */
+export interface StreamingAsr extends HybridObject<SpeechPlatforms> {
+    /** Optional callback invoked when a partial recognition result is available. */
+    onPartialResult?: (result: AsrResult) => void;
+    /** Optional callback invoked when a final recognition result is available. */
+    onFinalResult?: (result: AsrResult) => void;
+    /** Optional callback invoked on ASR errors. */
+    onError?: (error: string) => void;
+    load(config: AsrModelConfig): Promise<void>;
+    isLoaded(): boolean;
+    /** Feed streaming audio. */
+    acceptWaveform(samples: ArrayBuffer): Promise<void>;
+    /** Signal end of stream and return final result. */
+    finalize(): Promise<AsrResult>;
+    reset(): Promise<void>;
+    unload(): Promise<void>;
+}
+export type TtsModelType = "kokoro" | "vits" | "matcha" | "pocket" | "zipvoice";
+export interface TtsModelConfig {
+    type: TtsModelType;
+    /** Directory containing all model files. */
+    modelDir: string;
+    /**
+     * Single ONNX model file for Kokoro / VITS / ZipVoice.
+     * For Matcha, use acousticModel + vocoder instead.
+     */
+    model?: string;
+    /** Acoustic model ONNX file (Matcha only). */
+    acousticModel?: string;
+    /** Vocoder ONNX file (Matcha only). */
+    vocoder?: string;
+    /** Tokens / lexicon files. */
+    tokens?: string;
+    lexicon?: string;
+    /** Kokoro / Matcha voice data (e.g. voices.bin). */
+    voices?: string;
+    /** Optional espeak-ng-data directory (Kokoro / VITS / Matcha / ZipVoice). */
+    espeakNgData?: string;
+    /** Optional dict directory (Kokoro / VITS / Matcha legacy field). */
+    dictDir?: string;
+    /** Pocket: lm_main ONNX file. */
+    lmMain?: string;
+    /** Pocket: lm_flow ONNX file. */
+    lmFlow?: string;
+    /** Pocket: text conditioner ONNX file. */
+    textConditioner?: string;
+    /** Pocket: encoder ONNX file. */
+    pocketEncoder?: string;
+    /** Pocket: decoder ONNX file. */
+    pocketDecoder?: string;
+    /** Pocket: vocab.json file. */
+    vocabJson?: string;
+    /** Pocket: token_scores.json file. */
+    tokenScoresJson?: string;
+    /** ZipVoice: encoder ONNX file. */
+    zipvoiceEncoder?: string;
+    /** ZipVoice: decoder ONNX file. */
+    zipvoiceDecoder?: string;
+    /** Pocket: config JSON. */
+    config?: string;
+    /** Number of ONNX Runtime threads. Default: 2. */
+    numThreads?: number;
+    /** Output sample rate (some models produce 24 kHz). Default: 16000. */
+    outputSampleRate?: number;
+    /** Speaker ID for multi-speaker models. */
+    speakerId?: number;
+    /** Speed factor, e.g. 1.0. */
+    speed?: number;
+}
+export interface TtsResult {
+    /** Synthesized audio as 16 kHz mono f32 PCM (or configured output sample rate). */
+    samples: ArrayBuffer;
+    sampleRate: number;
+    /** Duration in milliseconds. */
+    durationMs: number;
+}
+/** Text-to-speech hybrid object. */
+export interface Tts extends HybridObject<SpeechPlatforms> {
+    load(config: TtsModelConfig): Promise<void>;
+    isLoaded(): boolean;
+    /** Synthesize text into audio on a background thread. Optional speed override. */
+    synthesize(text: string, speed?: number): Promise<TtsResult>;
+    /** Synthesize with a cloned speaker embedding (reference-audio cloning). */
+    synthesizeWithSpeaker(text: string, speakerId: string, speed?: number): Promise<TtsResult>;
+    /** Save TTS result as a WAV file at the given path. */
+    saveWav(result: TtsResult, path: string): Promise<void>;
+    unload(): Promise<void>;
+}
+export interface SpeakerEmbeddingConfig {
+    /** Directory containing the speaker embedding model. */
+    modelDir: string;
+    /** Model ONNX file. */
+    model: string;
+    numThreads: number;
+}
+export interface RegisteredSpeaker {
+    /** Stable speaker ID. */
+    id: string;
+    /** Optional display name. */
+    name: string;
+    /** Stored embedding file path on native side. */
+    embeddingPath: string;
+}
+/** Speaker embedding and voice cloning hybrid object. */
+export interface SpeakerManager extends HybridObject<SpeechPlatforms> {
+    load(config: SpeakerEmbeddingConfig): Promise<void>;
+    isLoaded(): boolean;
+    /** Compute an embedding from reference 16 kHz mono f32 PCM audio. */
+    computeEmbedding(samples: ArrayBuffer): Promise<ArrayBuffer>;
+    /** Register a speaker embedding for later TTS use. */
+    registerSpeaker(id: string, name: string, embedding: ArrayBuffer): Promise<RegisteredSpeaker>;
+    /** Register from a reference audio file. */
+    registerSpeakerFromFile(id: string, name: string, path: string): Promise<RegisteredSpeaker>;
+    /** List registered speakers. */
+    listSpeakers(): Promise<RegisteredSpeaker[]>;
+    /** Remove a registered speaker. */
+    removeSpeaker(id: string): Promise<void>;
+    unload(): Promise<void>;
+}
+export interface OnnxSpeech extends HybridObject<SpeechPlatforms> {
+    /** Create a VAD instance. */
+    createVad(): Vad;
+    /** Create an offline ASR instance. */
+    createOfflineAsr(): OfflineAsr;
+    /** Create a streaming ASR instance. */
+    createStreamingAsr(): StreamingAsr;
+    /** Create a TTS instance. */
+    createTts(): Tts;
+    /** Create a speaker manager for voice cloning. */
+    createSpeakerManager(): SpeakerManager;
+    /** Get the module version. */
+    readonly version: string;
+    /** Check if the CPU is a Qualcomm chip. Always returns false on iOS. */
+    isQualcommCpu(): boolean;
+}
+//# sourceMappingURL=OnnxSpeech.nitro.d.ts.map

package/lib/specs/OnnxSpeech.nitro.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"OnnxSpeech.nitro.d.ts","sourceRoot":"","sources":["../../src/specs/OnnxSpeech.nitro.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,4BAA4B,CAAC;AAE/D,0EAA0E;AAC1E,MAAM,MAAM,eAAe,GAAG;IAAE,GAAG,EAAE,KAAK,CAAC;IAAC,OAAO,EAAE,KAAK,CAAA;CAAE,CAAC;AAM7D,qDAAqD;AACrD,MAAM,WAAW,WAAW;IAC1B,uDAAuD;IACvD,UAAU,EAAE,MAAM,CAAC;IACnB,2CAA2C;IAC3C,QAAQ,EAAE,MAAM,CAAC;IACjB,yBAAyB;IACzB,YAAY,EAAE,OAAO,GAAG,OAAO,CAAC;CACjC;AAED,8EAA8E;AAC9E,MAAM,WAAW,MAAM,CAAC,CAAC;IACvB,OAAO,EAAE,OAAO,CAAC;IACjB,IAAI,CAAC,EAAE,CAAC,CAAC;IACT,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAMD,MAAM,WAAW,SAAS;IACxB;;;OAGG;IACH,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,sEAAsE;IACtE,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,iFAAiF;IACjF,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B,6DAA6D;IAC7D,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B,iFAAiF;IACjF,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,6DAA6D;IAC7D,OAAO,EAAE,MAAM,CAAC;IAChB,kCAAkC;IAClC,KAAK,EAAE,MAAM,CAAC;IACd,4CAA4C;IAC5C,OAAO,EAAE,WAAW,CAAC;CACtB;AAED,iDAAiD;AACjD,MAAM,WAAW,SAAS;IACxB,aAAa,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IAC9C,WAAW,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IAC5C,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,IAAI,CAAC;CACnC;AAED,8CAA8C;AAC9C,MAAM,WAAW,GAAI,SAAQ,YAAY,CAAC,eAAe,CAAC;IACxD,oDAAoD;IACpD,aAAa,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IAC9C,kDAAkD;IAClD,WAAW,CAAC,EAAE,CAAC,OAAO,EAAE,UAAU,KAAK,IAAI,CAAC;IAC5C,+CAA+C;IAC/C,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,IAAI,CAAC;IAClC,gEAAgE;IAChE,UAAU,CAAC,MAAM,EAAE,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IAC7C,4DAA4D;IAC5D,aAAa,IAAI,OAAO,CAAC;IACzB,8EAA8E;IAC9E,OAAO,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IAC7C,0EAA0E;IAC1E,YAAY,IAAI,OAAO,CAAC,UAAU,EAAE,CAAC,CAAC;IACtC,wCAAwC;IACxC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACxB;AAMD,MAAM,MAAM,YAAY,GACpB,SAAS,GACT,YAAY,GACZ,YAAY,GACZ,WAAW,GACX,WAAW,GACX,OAAO,GACP,YAAY,GACZ,WAAW,GACX,SAAS,GACT,MAAM,GACN,aAAa,CAAC;AAElB,MAAM,WAAW,cAAc;IAC7B,IAAI,EAAE,YAAY,CAAC;IACnB,4CAA4C;IAC5C,QAAQ,EAAE,MAAM,CAAC;IACjB,wDAAwD;IACxD,UAAU,EAAE,MAAM,CAAC;IACnB,6CAA6C;IAC7C,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,kEAAkE;IAClE,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,uEAAuE;IACvE,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,gCAAgC;IAChC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,sDAAsD;IACtD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,qEAAqE;IACrE,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,wCAAwC;IACxC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,8CAA8C;IAC9C,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,sCAAsC;IACtC,MAAM,CAAC,EAAE,OAAO,CAAC;CAClB;AAED,MAAM,WAAW,SAAS;IACxB,IAAI,EAAE,MAAM,CAAC;IACb,uCAAuC;IACvC,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,oCAAoC;IACpC,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,kCAAkC;IAClC,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,2DAA2D;IAC3D,UAAU,CAAC,EAAE,MAAM,EAAE,CAAC;IACtB,6CAA6C;IAC7C,IAAI,CAAC,EAAE,MAAM,CAAC;CACf;AAED,MAAM,WAAW,kBAAkB;IACjC,eAAe,CAAC,EAAE,CAAC,MAAM,EAAE,SAAS,KAAK,IAAI,CAAC;IAC9C,aAAa,CAAC,EAAE,CAAC,MAAM,EAAE,SAAS,KAAK,IAAI,CAAC;IAC5C,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,IAAI,CAAC;CACnC;AAED,8CAA8C;AAC9C,MAAM,WAAW,UAAW,SAAQ,YAAY,CAAC,eAAe,CAAC;IAC/D,mCAAmC;IACnC,IAAI,CAAC,MAAM,EAAE,cAAc,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IAC5C,QAAQ,IAAI,OAAO,CAAC;IACpB,sDAAsD;IACtD,SAAS,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC;IACpD,gEAAgE;IAChE,aAAa,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC;IAChD,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACzB;AAED,mCAAmC;AACnC,MAAM,WAAW,YAAa,SAAQ,YAAY,CAAC,eAAe,CAAC;IACjE,gFAAgF;IAChF,eAAe,CAAC,EAAE,CAAC,MAAM,EAAE,SAAS,KAAK,IAAI,CAAC;IAC9C,8EAA8E;IAC9E,aAAa,CAAC,EAAE,CAAC,MAAM,EAAE,SAAS,KAAK,IAAI,CAAC;IAC5C,+CAA+C;IAC/C,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,IAAI,CAAC;IAClC,IAAI,CAAC,MAAM,EAAE,cAAc,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IAC5C,QAAQ,IAAI,OAAO,CAAC;IACpB,4BAA4B;IAC5B,cAAc,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACpD,oDAAoD;IACpD,QAAQ,IAAI,OAAO,CAAC,SAAS,CAAC,CAAC;IAC/B,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IACvB,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACzB;AAMD,MAAM,MAAM,YAAY,GACpB,QAAQ,GACR,MAAM,GACN,QAAQ,GACR,QAAQ,GACR,UAAU,CAAC;AAEf,MAAM,WAAW,cAAc;IAC7B,IAAI,EAAE,YAAY,CAAC;IACnB,4CAA4C;IAC5C,QAAQ,EAAE,MAAM,CAAC;IACjB;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,8CAA8C;IAC9C,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,uCAAuC;IACvC,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,8BAA8B;IAC9B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,oDAAoD;IACpD,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,6EAA6E;IAC7E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,qEAAqE;IACrE,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,iCAAiC;IACjC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,iCAAiC;IACjC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,0CAA0C;IAC1C,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,iCAAiC;IACjC,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,iCAAiC;IACjC,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,+BAA+B;IAC/B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,sCAAsC;IACtC,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,mCAAmC;IACnC,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,mCAAmC;IACnC,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,2BAA2B;IAC3B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,kDAAkD;IAClD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,uEAAuE;IACvE,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,2CAA2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,8BAA8B;IAC9B,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,SAAS;IACxB,mFAAmF;IACnF,OAAO,EAAE,WAAW,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,gCAAgC;IAChC,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,oCAAoC;AACpC,MAAM,WAAW,GAAI,SAAQ,YAAY,CAAC,eAAe,CAAC;IACxD,IAAI,CAAC,MAAM,EAAE,cAAc,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IAC5C,QAAQ,IAAI,OAAO,CAAC;IACpB,kFAAkF;IAClF,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC;IAC7D,4EAA4E;IAC5E,qBAAqB,CAAC,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,KAAK,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC;IAC3F,uDAAuD;IACvD,OAAO,CAAC,MAAM,EAAE,SAAS,EAAE,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACxD,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACzB;AAMD,MAAM,WAAW,sBAAsB;IACrC,wDAAwD;IACxD,QAAQ,EAAE,MAAM,CAAC;IACjB,uBAAuB;IACvB,KAAK,EAAE,MAAM,CAAC;IACd,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,iBAAiB;IAChC,yBAAyB;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,6BAA6B;IAC7B,IAAI,EAAE,MAAM,CAAC;IACb,iDAAiD;IACjD,aAAa,EAAE,MAAM,CAAC;CACvB;AAED,yDAAyD;AACzD,MAAM,WAAW,cAAe,SAAQ,YAAY,CAAC,eAAe,CAAC;IACnE,IAAI,CAAC,MAAM,EAAE,sBAAsB,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACpD,QAAQ,IAAI,OAAO,CAAC;IACpB,qEAAqE;IACrE,gBAAgB,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC,CAAC;IAC7D,sDAAsD;IACtD,eAAe,CAAC,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,WAAW,GAAG,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC9F,4CAA4C;IAC5C,uBAAuB,CAAC,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAC5F,gCAAgC;IAChC,YAAY,IAAI,OAAO,CAAC,iBAAiB,EAAE,CAAC,CAAC;IAC7C,mCAAmC;IACnC,aAAa,CAAC,EAAE,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACzC,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACzB;AAMD,MAAM,WAAW,UAAW,SAAQ,YAAY,CAAC,eAAe,CAAC;IAC/D,6BAA6B;IAC7B,SAAS,IAAI,GAAG,CAAC;IACjB,sCAAsC;IACtC,gBAAgB,IAAI,UAAU,CAAC;IAC/B,uCAAuC;IACvC,kBAAkB,IAAI,YAAY,CAAC;IACnC,6BAA6B;IAC7B,SAAS,IAAI,GAAG,CAAC;IACjB,kDAAkD;IAClD,oBAAoB,IAAI,cAAc,CAAC;IACvC,8BAA8B;IAC9B,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;IACzB,wEAAwE;IACxE,aAAa,IAAI,OAAO,CAAC;CAC1B"}

package/lib/specs/OnnxSpeech.nitro.js

@@ -0,0 +1,6 @@
+// ------------------------------------------------------------------------------
+// Nitro TypeScript specifications for react-native-nitro-onnx.
+// All numeric audio data is 16 kHz mono PCM f32 unless otherwise noted.
+// ------------------------------------------------------------------------------
+export {};
+//# sourceMappingURL=OnnxSpeech.nitro.js.map

package/lib/specs/OnnxSpeech.nitro.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"OnnxSpeech.nitro.js","sourceRoot":"","sources":["../../src/specs/OnnxSpeech.nitro.ts"],"names":[],"mappings":"AAAA,iFAAiF;AACjF,+DAA+D;AAC/D,wEAAwE;AACxE,iFAAiF"}

package/nitro.json

@@ -0,0 +1,19 @@
+{
+  "cxxNamespace": ["onnx", "speech"],
+  "ios": {
+    "iosModuleName": "NitroOnnxSpeech"
+  },
+  "android": {
+    "androidNamespace": ["onnx", "speech"],
+    "androidCxxLibName": "NitroOnnxSpeech"
+  },
+  "autolinking": {
+    "NitroOnnxSpeech": {
+      "all": {
+        "language": "c++",
+        "implementationClassName": "NitroOnnxSpeech"
+      }
+    }
+  },
+  "ignorePaths": ["node_modules", "lib"]
+}

package/nitrogen/generated/.gitattributes

@@ -0,0 +1 @@
+** linguist-generated=true

package/nitrogen/generated/android/NitroOnnxSpeech+autolinking.cmake

@@ -0,0 +1,86 @@
+#
+# NitroOnnxSpeech+autolinking.cmake
+# This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+# https://github.com/mrousavy/nitro
+# Copyright © Marc Rousavy @ Margelo
+#
+
+# This is a CMake file that adds all files generated by Nitrogen
+# to the current CMake project.
+#
+# To use it, add this to your CMakeLists.txt:
+# ```cmake
+# include(${CMAKE_SOURCE_DIR}/../nitrogen/generated/android/NitroOnnxSpeech+autolinking.cmake)
+# ```
+
+# Define a flag to check if we are building properly
+add_definitions(-DBUILDING_NITROONNXSPEECH_WITH_GENERATED_CMAKE_PROJECT)
+
+# Enable Raw Props parsing in react-native (for Nitro Views)
+add_definitions(-DRN_SERIALIZABLE_STATE)
+
+# Add all headers that were generated by Nitrogen
+include_directories(
+  "../nitrogen/generated/shared/c++"
+  "../nitrogen/generated/android/c++"
+  "../nitrogen/generated/android/"
+)
+
+# Add all .cpp sources that were generated by Nitrogen
+target_sources(
+  # CMake project name (Android C++ library name)
+  NitroOnnxSpeech PRIVATE
+  # Autolinking Setup
+  ../nitrogen/generated/android/NitroOnnxSpeechOnLoad.cpp
+  # Shared Nitrogen C++ sources
+  ../nitrogen/generated/shared/c++/HybridVadSpec.cpp
+  ../nitrogen/generated/shared/c++/HybridOfflineAsrSpec.cpp
+  ../nitrogen/generated/shared/c++/HybridStreamingAsrSpec.cpp
+  ../nitrogen/generated/shared/c++/HybridTtsSpec.cpp
+  ../nitrogen/generated/shared/c++/HybridSpeakerManagerSpec.cpp
+  ../nitrogen/generated/shared/c++/HybridOnnxSpeechSpec.cpp
+  # Android-specific Nitrogen C++ sources
+  
+)
+
+# From node_modules/react-native/ReactAndroid/cmake-utils/folly-flags.cmake
+# Used in node_modules/react-native/ReactAndroid/cmake-utils/ReactNative-application.cmake
+target_compile_definitions(
+  NitroOnnxSpeech PRIVATE
+  -DFOLLY_NO_CONFIG=1
+  -DFOLLY_HAVE_CLOCK_GETTIME=1
+  -DFOLLY_USE_LIBCPP=1
+  -DFOLLY_CFG_NO_COROUTINES=1
+  -DFOLLY_MOBILE=1
+  -DFOLLY_HAVE_RECVMMSG=1
+  -DFOLLY_HAVE_PTHREAD=1
+  # Once we target android-23 above, we can comment
+  # the following line. NDK uses GNU style stderror_r() after API 23.
+  -DFOLLY_HAVE_XSI_STRERROR_R=1
+)
+
+# Add all libraries required by the generated specs
+find_package(fbjni REQUIRED) # <-- Used for communication between Java <-> C++
+find_package(ReactAndroid REQUIRED) # <-- Used to set up React Native bindings (e.g. CallInvoker/TurboModule)
+find_package(react-native-nitro-modules REQUIRED) # <-- Used to create all HybridObjects and use the Nitro core library
+
+# Link all libraries together
+target_link_libraries(
+        NitroOnnxSpeech
+        fbjni::fbjni                              # <-- Facebook C++ JNI helpers
+        ReactAndroid::jsi                         # <-- RN: JSI
+        react-native-nitro-modules::NitroModules  # <-- NitroModules Core :)
+)
+
+# Link react-native (different prefab between RN 0.75 and RN 0.76)
+if(ReactAndroid_VERSION_MINOR GREATER_EQUAL 76)
+    target_link_libraries(
+        NitroOnnxSpeech
+        ReactAndroid::reactnative                 # <-- RN: Native Modules umbrella prefab
+    )
+else()
+    target_link_libraries(
+        NitroOnnxSpeech
+        ReactAndroid::react_nativemodule_core     # <-- RN: TurboModules Core
+    )
+endif()

package/nitrogen/generated/android/NitroOnnxSpeech+autolinking.gradle

@@ -0,0 +1,27 @@
+///
+/// NitroOnnxSpeech+autolinking.gradle
+/// This file was generated by nitrogen. DO NOT MODIFY THIS FILE.
+/// https://github.com/mrousavy/nitro
+/// Copyright © Marc Rousavy @ Margelo
+///
+
+/// This is a Gradle file that adds all files generated by Nitrogen
+/// to the current Gradle project.
+///
+/// To use it, add this to your build.gradle:
+/// ```gradle
+/// apply from: '../nitrogen/generated/android/NitroOnnxSpeech+autolinking.gradle'
+/// ```
+
+logger.warn("[NitroModules] 🔥 NitroOnnxSpeech is boosted by nitro!")
+
+android {
+  sourceSets {
+    main {
+      java.srcDirs += [
+        // Nitrogen files
+        "${project.projectDir}/../nitrogen/generated/android/kotlin"
+      ]
+    }
+  }
+}