whisper.rn 0.1.3 → 0.1.4

package/android/src/main/java/com/rnwhisper/RNWhisperModule.java

@@ -12,6 +12,7 @@ import com.facebook.react.bridge.ReactContextBaseJavaModule;
 import com.facebook.react.bridge.ReactMethod;
 import com.facebook.react.bridge.LifecycleEventListener;
 import com.facebook.react.bridge.ReadableMap;
+import com.facebook.react.bridge.WritableMap;
 import com.facebook.react.module.annotations.ReactModule;
 
 import java.util.HashMap;
@@ -72,11 +73,11 @@ public class RNWhisperModule extends ReactContextBaseJavaModule implements Lifec
 
   @ReactMethod
   public void transcribe(int id, String filePath, ReadableMap options, Promise promise) {
-    new AsyncTask<Void, Void, String>() {
+    new AsyncTask<Void, Void, WritableMap>() {
       private Exception exception;
 
       @Override
-      protected String doInBackground(Void... voids) {
+      protected WritableMap doInBackground(Void... voids) {
         try {
           WhisperContext context = contexts.get(id);
           if (context == null) {
@@ -90,12 +91,12 @@ public class RNWhisperModule extends ReactContextBaseJavaModule implements Lifec
       }
 
       @Override
-      protected void onPostExecute(String result) {
+      protected void onPostExecute(WritableMap data) {
         if (exception != null) {
           promise.reject(exception);
           return;
         }
-        promise.resolve(result);
+        promise.resolve(data);
       }
     }.execute();
   }

package/android/src/main/java/com/rnwhisper/WhisperContext.java

@@ -1,5 +1,8 @@
 package com.rnwhisper;
 
+import com.facebook.react.bridge.Arguments;
+import com.facebook.react.bridge.WritableArray;
+import com.facebook.react.bridge.WritableMap;
 import com.facebook.react.bridge.ReadableMap;
 
 import android.util.Log;
@@ -29,7 +32,7 @@ public class WhisperContext {
     this.context = context;
   }
 
-  public String transcribe(final String filePath, final ReadableMap options) throws IOException, Exception {
+  public WritableMap transcribe(final String filePath, final ReadableMap options) throws IOException, Exception {
     int code = fullTranscribe(
       context,
       decodeWaveFile(new File(filePath)),
@@ -37,14 +40,18 @@ public class WhisperContext {
       options.hasKey("maxThreads") ? options.getInt("maxThreads") : -1,
       // jint max_context,
       options.hasKey("maxContext") ? options.getInt("maxContext") : -1,
+
+      // jint word_thold,
+      options.hasKey("wordThold") ? options.getInt("wordThold") : -1,
       // jint max_len,
       options.hasKey("maxLen") ? options.getInt("maxLen") : -1,
+      // jboolean token_timestamps,
+      options.hasKey("tokenTimestamps") ? options.getBoolean("tokenTimestamps") : false,
+  
       // jint offset,
       options.hasKey("offset") ? options.getInt("offset") : -1,
       // jint duration,
       options.hasKey("duration") ? options.getInt("duration") : -1,
-      // jint word_thold,
-      options.hasKey("wordThold") ? options.getInt("wordThold") : -1,
       // jfloat temperature,
       options.hasKey("temperature") ? (float) options.getDouble("temperature") : -1.0f,
       // jfloat temperature_inc,
@@ -58,17 +65,31 @@ public class WhisperContext {
       // jboolean translate,
       options.hasKey("translate") ? options.getBoolean("translate") : false,
       // jstring language,
-      options.hasKey("language") ? options.getString("language") : "auto"
+      options.hasKey("language") ? options.getString("language") : "auto",
+      // jstring prompt
+      options.hasKey("prompt") ? options.getString("prompt") : null
     );
     if (code != 0) {
       throw new Exception("Transcription failed with code " + code);
     }
     Integer count = getTextSegmentCount(context);
     StringBuilder builder = new StringBuilder();
+
+    WritableMap data = Arguments.createMap();
+    WritableArray segments = Arguments.createArray();
     for (int i = 0; i < count; i++) {
-      builder.append(getTextSegment(context, i));
+      String text = getTextSegment(context, i);
+      builder.append(text);
+
+      WritableMap segment = Arguments.createMap();
+      segment.putString("text", text);
+      segment.putInt("t0", getTextSegmentT0(context, i));
+      segment.putInt("t1", getTextSegmentT1(context, i));
+      segments.pushMap(segment);
     }
-    return builder.toString();
+    data.putString("result", builder.toString());
+    data.putArray("segments", segments);
+    return data;
   }
 
   public void release() {
@@ -168,19 +189,23 @@ public class WhisperContext {
     float[] audio_data,
     int n_threads,
     int max_context,
+    int word_thold,
     int max_len,
+    boolean token_timestamps,
     int offset,
     int duration,
-    int word_thold,
     float temperature,
     float temperature_inc,
     int beam_size,
     int best_of,
     boolean speed_up,
     boolean translate,
-    String language
+    String language,
+    String prompt
   );
   protected static native int getTextSegmentCount(long context);
   protected static native String getTextSegment(long context, int index);
+  protected static native int getTextSegmentT0(long context, int index);
+  protected static native int getTextSegmentT1(long context, int index);
   protected static native void freeContext(long contextPtr);
 }

package/android/src/main/jni/whisper/Whisper.mk

@@ -15,4 +15,5 @@ LOCAL_CFLAGS    += -DSTDC_HEADERS -std=c11 -I $(WHISPER_LIB_DIR)
 LOCAL_CPPFLAGS  += -std=c++11
 LOCAL_SRC_FILES := $(WHISPER_LIB_DIR)/ggml.c \
                    $(WHISPER_LIB_DIR)/whisper.cpp \
-                   $(LOCAL_PATH)/jni.c
+                   $(WHISPER_LIB_DIR)/rn-whisper.cpp \
+                   $(LOCAL_PATH)/jni.cpp

package/android/src/main/jni/whisper/{jni.c → jni.cpp}

@@ -2,10 +2,11 @@
 #include <android/asset_manager.h>
 #include <android/asset_manager_jni.h>
 #include <android/log.h>
-#include <stdlib.h>
+#include <cstdlib>
 #include <sys/sysinfo.h>
-#include <string.h>
+#include <string>
 #include "whisper.h"
+#include "rn-whisper.h"
 #include "ggml.h"
 
 #define UNUSED(x) (void)(x)
@@ -22,27 +23,17 @@ static inline int max(int a, int b) {
     return (a > b) ? a : b;
 }
 
-static size_t asset_read(void *ctx, void *output, size_t read_size) {
-    return AAsset_read((AAsset *) ctx, output, read_size);
-}
-
-static bool asset_is_eof(void *ctx) {
-    return AAsset_getRemainingLength64((AAsset *) ctx) <= 0;
-}
-
-static void asset_close(void *ctx) {
-    AAsset_close((AAsset *) ctx);
-}
+extern "C" {
 
 JNIEXPORT jlong JNICALL
 Java_com_rnwhisper_WhisperContext_initContext(
         JNIEnv *env, jobject thiz, jstring model_path_str) {
     UNUSED(thiz);
-    struct whisper_context *context = NULL;
-    const char *model_path_chars = (*env)->GetStringUTFChars(env, model_path_str, NULL);
+    struct whisper_context *context = nullptr;
+    const char *model_path_chars = env->GetStringUTFChars(model_path_str, nullptr);
     context = whisper_init_from_file(model_path_chars);
-    (*env)->ReleaseStringUTFChars(env, model_path_str, model_path_chars);
-    return (jlong) context;
+    env->ReleaseStringUTFChars(model_path_str, model_path_chars);
+    return reinterpret_cast<jlong>(context);
 }
 
 JNIEXPORT jint JNICALL
@@ -53,29 +44,31 @@ Java_com_rnwhisper_WhisperContext_fullTranscribe(
     jfloatArray audio_data,
     jint n_threads,
     jint max_context,
-    jint max_len,
+    int word_thold,
+    int max_len,
+    jboolean token_timestamps,
     jint offset,
     jint duration,
-    jint word_thold,
     jfloat temperature,
     jfloat temperature_inc,
     jint beam_size,
     jint best_of,
     jboolean speed_up,
     jboolean translate,
-    jstring language
+    jstring language,
+    jstring prompt
 ) {
     UNUSED(thiz);
-    struct whisper_context *context = (struct whisper_context *) context_ptr;
-    jfloat *audio_data_arr = (*env)->GetFloatArrayElements(env, audio_data, NULL);
-    const jsize audio_data_length = (*env)->GetArrayLength(env, audio_data);
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
+    jfloat *audio_data_arr = env->GetFloatArrayElements(audio_data, nullptr);
+    const jsize audio_data_length = env->GetArrayLength(audio_data);
 
     int max_threads = max(1, min(8, get_nprocs() - 2));
 
     LOGI("About to create params");
 
     struct whisper_full_params params = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
-    
+
     if (beam_size > -1) {
         params.strategy = WHISPER_SAMPLING_BEAM_SEARCH;
         params.beam_search.beam_size = beam_size;
@@ -86,22 +79,25 @@ Java_com_rnwhisper_WhisperContext_fullTranscribe(
     params.print_timestamps = false;
     params.print_special = false;
     params.translate = translate;
-    params.language = language;
+    const char *language_chars = env->GetStringUTFChars(language, nullptr);
+    params.language = language_chars;
     params.n_threads = n_threads > 0 ? n_threads : max_threads;
     params.speed_up = speed_up;
     params.offset_ms = 0;
     params.no_context = true;
     params.single_segment = false;
 
+    if (max_len > -1) {
+        params.max_len = max_len;
+    }
+    params.token_timestamps = token_timestamps;
+
     if (best_of > -1) {
         params.greedy.best_of = best_of;
     }
     if (max_context > -1) {
         params.n_max_text_ctx = max_context;
     }
-    if (max_len > -1) {
-        params.max_len = max_len;
-    }
     if (offset > -1) {
         params.offset_ms = offset;
     }
@@ -117,6 +113,13 @@ Java_com_rnwhisper_WhisperContext_fullTranscribe(
     if (temperature_inc > -1) {
         params.temperature_inc = temperature_inc;
     }
+    if (prompt != nullptr) {
+        rn_whisper_convert_prompt(
+            context,
+            params,
+            new std::string(env->GetStringUTFChars(prompt, nullptr))
+        );
+    }
 
     LOGI("About to reset timings");
     whisper_reset_timings(context);
@@ -126,7 +129,8 @@ Java_com_rnwhisper_WhisperContext_fullTranscribe(
     if (code == 0) {
         // whisper_print_timings(context);
     }
-    (*env)->ReleaseFloatArrayElements(env, audio_data, audio_data_arr, JNI_ABORT);
+    env->ReleaseFloatArrayElements(audio_data, audio_data_arr, JNI_ABORT);
+    env->ReleaseStringUTFChars(language, language_chars);
     return code;
 }
 
@@ -135,7 +139,7 @@ Java_com_rnwhisper_WhisperContext_getTextSegmentCount(
         JNIEnv *env, jobject thiz, jlong context_ptr) {
     UNUSED(env);
     UNUSED(thiz);
-    struct whisper_context *context = (struct whisper_context *) context_ptr;
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
     return whisper_full_n_segments(context);
 }
 
@@ -143,17 +147,37 @@ JNIEXPORT jstring JNICALL
 Java_com_rnwhisper_WhisperContext_getTextSegment(
         JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
     UNUSED(thiz);
-    struct whisper_context *context = (struct whisper_context *) context_ptr;
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
     const char *text = whisper_full_get_segment_text(context, index);
-    jstring string = (*env)->NewStringUTF(env, text);
+    jstring string = env->NewStringUTF(text);
     return string;
 }
 
+JNIEXPORT jint JNICALL
+Java_com_rnwhisper_WhisperContext_getTextSegmentT0(
+        JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
+    UNUSED(env);
+    UNUSED(thiz);
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
+    return whisper_full_get_segment_t0(context, index);
+}
+
+JNIEXPORT jint JNICALL
+Java_com_rnwhisper_WhisperContext_getTextSegmentT1(
+        JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
+    UNUSED(env);
+    UNUSED(thiz);
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
+    return whisper_full_get_segment_t1(context, index);
+}
+
 JNIEXPORT void JNICALL
 Java_com_rnwhisper_WhisperContext_freeContext(
         JNIEnv *env, jobject thiz, jlong context_ptr) {
     UNUSED(env);
     UNUSED(thiz);
-    struct whisper_context *context = (struct whisper_context *) context_ptr;
+    struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
     whisper_free(context);
 }
+
+} // extern "C"

package/cpp/rn-whisper.cpp

@@ -0,0 +1,31 @@
+#include <cstdio>
+#include <string>
+#include <vector>
+#include "whisper.h"
+
+extern "C" {
+
+void rn_whisper_convert_prompt(
+  struct whisper_context * ctx,
+  struct whisper_full_params params,
+  std::string * prompt
+) {
+  std::vector<whisper_token> prompt_tokens;
+  if (!prompt->empty()) {
+    prompt_tokens.resize(1024);
+    prompt_tokens.resize(whisper_tokenize(ctx, prompt->c_str(), prompt_tokens.data(), prompt_tokens.size()));
+
+    // fprintf(stderr, "\n");
+    // fprintf(stderr, "initial prompt: '%s'\n", prompt->c_str());
+    // fprintf(stderr, "initial tokens: [ ");
+    // for (int i = 0; i < (int) prompt_tokens.size(); ++i) {
+    //     fprintf(stderr, "%d ", prompt_tokens[i]);
+    // }
+    // fprintf(stderr, "]\n");
+
+    params.prompt_tokens = prompt_tokens.data();
+    params.prompt_n_tokens = prompt_tokens.size();
+  }
+}
+
+}

package/cpp/rn-whisper.h

@@ -0,0 +1,16 @@
+
+#ifdef __cplusplus
+#include <string>
+#include <whisper.h>
+extern "C" {
+#endif
+
+void rn_whisper_convert_prompt(
+  struct whisper_context * ctx,
+  struct whisper_full_params params,
+  std::string * prompt
+);
+
+#ifdef __cplusplus
+}
+#endif

package/ios/RNWhisper.h

@@ -1,7 +1,9 @@
 #ifdef __cplusplus
 #import "whisper.h"
+#import "rn-whisper.h"
 #endif
 
+
 #import <React/RCTBridgeModule.h>
 
 @interface RNWhisper : NSObject <RCTBridgeModule>

package/ios/RNWhisper.mm

@@ -1,6 +1,7 @@
 
 #import "RNWhisper.h"
 #include <stdlib.h>
+#include <string>
 
 @interface WhisperContext : NSObject {
 }
@@ -93,15 +94,18 @@ RCT_REMAP_METHOD(transcribe,
     params.no_context       = true;
     params.single_segment   = false;
 
+    if (options[@"maxLen"] != nil) {
+        params.max_len = [options[@"maxLen"] intValue];
+    }
+    params.token_timestamps = options[@"tokenTimestamps"] != nil ? [options[@"tokenTimestamps"] boolValue] : false;
+
     if (options[@"bestOf"] != nil) {
         params.greedy.best_of = [options[@"bestOf"] intValue];
     }
     if (options[@"maxContext"] != nil) {
         params.n_max_text_ctx = [options[@"maxContext"] intValue];
     }
-    if (options[@"maxLen"] != nil) {
-        params.max_len = [options[@"maxLen"] intValue];
-    }
+    
     if (options[@"offset"] != nil) {
         params.offset_ms = [options[@"offset"] intValue];
     }
@@ -117,6 +121,15 @@ RCT_REMAP_METHOD(transcribe,
     if (options[@"temperatureInc"] != nil) {
         params.temperature_inc = [options[@"temperature_inc"] floatValue];
     }
+    
+    if (options[@"prompt"] != nil) {
+        std::string *prompt = new std::string([options[@"prompt"] UTF8String]);
+        rn_whisper_convert_prompt(
+            context.ctx,
+            params,
+            prompt
+        );
+    }
 
     whisper_reset_timings(context.ctx);
     int code = whisper_full(context.ctx, params, waveFile, count);
@@ -132,11 +145,25 @@ RCT_REMAP_METHOD(transcribe,
 
     NSString *result = @"";
     int n_segments = whisper_full_n_segments(context.ctx);
+
+    NSMutableArray *segments = [[NSMutableArray alloc] init];
     for (int i = 0; i < n_segments; i++) {
         const char * text_cur = whisper_full_get_segment_text(context.ctx, i);
         result = [result stringByAppendingString:[NSString stringWithUTF8String:text_cur]];
-    }
-    resolve(result);
+
+        const int64_t t0 = whisper_full_get_segment_t0(context.ctx, i);
+        const int64_t t1 = whisper_full_get_segment_t1(context.ctx, i);
+        NSDictionary *segment = @{
+            @"text": [NSString stringWithUTF8String:text_cur],
+            @"t0": [NSNumber numberWithLongLong:t0],
+            @"t1": [NSNumber numberWithLongLong:t1]
+        };
+        [segments addObject:segment];
+    }
+    resolve(@{
+        @"result": result,
+        @"segments": segments
+    });
 }
 
 RCT_REMAP_METHOD(releaseContext,

package/jest/mock.js

@@ -3,7 +3,10 @@ const { NativeModules } = require('react-native')
 if (!NativeModules.RNWhisper) {
   NativeModules.RNWhisper = {
     initContext: jest.fn(() => Promise.resolve(1)),
-    transcribe: jest.fn(() => Promise.resolve('TEST')),
+    transcribe: jest.fn(() => Promise.resolve({
+      result: ' Test',
+      segments: [{ text: ' Test', t0: 0, t1: 33 }],
+    })),
     releaseContext: jest.fn(() => Promise.resolve()),
     releaseAllContexts: jest.fn(() => Promise.resolve()),
   }

package/lib/commonjs/index.js

@@ -21,9 +21,7 @@ class WhisperContext {
   }
   async transcribe(path) {
     let options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};
-    return RNWhisper.transcribe(this.id, path, options).then(result => ({
-      result
-    }));
+    return RNWhisper.transcribe(this.id, path, options);
   }
   async release() {
     return RNWhisper.releaseContext(this.id);

package/lib/commonjs/index.js.map

@@ -1 +1 @@
-{"version":3,"names":["_reactNative","require","LINKING_ERROR","Platform","select","ios","default","RNWhisper","NativeModules","Proxy","get","Error","WhisperContext","constructor","id","transcribe","path","options","arguments","length","undefined","then","result","release","releaseContext","initWhisper","filePath","initContext","releaseAllWhisper","releaseAllContexts"],"sourceRoot":"../../src","sources":["index.tsx"],"mappings":";;;;;;;AAAA,IAAAA,YAAA,GAAAC,OAAA;AAEA,MAAMC,aAAa,GAChB,sEAAqEC,qBAAQ,CAACC,MAAM,CAAC;EAAEC,GAAG,EAAE,gCAAgC;EAAEC,OAAO,EAAE;AAAG,CAAC,CAC3I,oDAAmD;AAEtD,MAAMC,SAAS,GAAGC,0BAAa,CAACD,SAAS,GACrCC,0BAAa,CAACD,SAAS,GACvB,IAAIE,KAAK,CACT,CAAC,CAAC,EACF;EACEC,GAAGA,CAAA,EAAG;IACJ,MAAM,IAAIC,KAAK,CAACT,aAAa,CAAC;EAChC;AACF,CAAC,CACF;AAoBH,MAAMU,cAAc,CAAC;EAGnBC,WAAWA,CAACC,EAAU,EAAE;IACtB,IAAI,CAACA,EAAE,GAAGA,EAAE;EACd;EAEA,MAAMC,UAAUA,CAACC,IAAY,EAA8D;IAAA,IAA5DC,OAA0B,GAAAC,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;IAC5D,OAAOX,SAAS,CAACQ,UAAU,CAAC,IAAI,CAACD,EAAE,EAAEE,IAAI,EAAEC,OAAO,CAAC,CAACI,IAAI,CAAEC,MAAc,KAAM;MAC5EA;IACF,CAAC,CAAC,CAAC;EACL;EAEA,MAAMC,OAAOA,CAAA,EAAG;IACd,OAAOhB,SAAS,CAACiB,cAAc,CAAC,IAAI,CAACV,EAAE,CAAC;EAC1C;AACF;AAEO,eAAeW,WAAWA,CAAA,EAEN;EAAA,IADzB;IAAEC;EAAgC,CAAC,GAAAR,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;EAExC,MAAMJ,EAAE,GAAG,MAAMP,SAAS,CAACoB,WAAW,CAACD,QAAQ,CAAC;EAChD,OAAO,IAAId,cAAc,CAACE,EAAE,CAAC;AAC/B;AAEO,eAAec,iBAAiBA,CAAA,EAAkB;EACvD,OAAOrB,SAAS,CAACsB,kBAAkB,EAAE;AACvC"}
+{"version":3,"names":["_reactNative","require","LINKING_ERROR","Platform","select","ios","default","RNWhisper","NativeModules","Proxy","get","Error","WhisperContext","constructor","id","transcribe","path","options","arguments","length","undefined","release","releaseContext","initWhisper","filePath","initContext","releaseAllWhisper","releaseAllContexts"],"sourceRoot":"../../src","sources":["index.tsx"],"mappings":";;;;;;;AAAA,IAAAA,YAAA,GAAAC,OAAA;AAEA,MAAMC,aAAa,GAChB,sEAAqEC,qBAAQ,CAACC,MAAM,CAAC;EAAEC,GAAG,EAAE,gCAAgC;EAAEC,OAAO,EAAE;AAAG,CAAC,CAC3I,oDAAmD;AAEtD,MAAMC,SAAS,GAAGC,0BAAa,CAACD,SAAS,GACrCC,0BAAa,CAACD,SAAS,GACvB,IAAIE,KAAK,CACT,CAAC,CAAC,EACF;EACEC,GAAGA,CAAA,EAAG;IACJ,MAAM,IAAIC,KAAK,CAACT,aAAa,CAAC;EAChC;AACF,CAAC,CACF;AA6BH,MAAMU,cAAc,CAAC;EAGnBC,WAAWA,CAACC,EAAU,EAAE;IACtB,IAAI,CAACA,EAAE,GAAGA,EAAE;EACd;EAEA,MAAMC,UAAUA,CAACC,IAAY,EAA8D;IAAA,IAA5DC,OAA0B,GAAAC,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;IAC5D,OAAOX,SAAS,CAACQ,UAAU,CAAC,IAAI,CAACD,EAAE,EAAEE,IAAI,EAAEC,OAAO,CAAC;EACrD;EAEA,MAAMI,OAAOA,CAAA,EAAG;IACd,OAAOd,SAAS,CAACe,cAAc,CAAC,IAAI,CAACR,EAAE,CAAC;EAC1C;AACF;AAEO,eAAeS,WAAWA,CAAA,EAEN;EAAA,IADzB;IAAEC;EAAgC,CAAC,GAAAN,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;EAExC,MAAMJ,EAAE,GAAG,MAAMP,SAAS,CAACkB,WAAW,CAACD,QAAQ,CAAC;EAChD,OAAO,IAAIZ,cAAc,CAACE,EAAE,CAAC;AAC/B;AAEO,eAAeY,iBAAiBA,CAAA,EAAkB;EACvD,OAAOnB,SAAS,CAACoB,kBAAkB,EAAE;AACvC"}

package/lib/module/index.js

@@ -14,9 +14,7 @@ class WhisperContext {
   }
   async transcribe(path) {
     let options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};
-    return RNWhisper.transcribe(this.id, path, options).then(result => ({
-      result
-    }));
+    return RNWhisper.transcribe(this.id, path, options);
   }
   async release() {
     return RNWhisper.releaseContext(this.id);

package/lib/module/index.js.map

@@ -1 +1 @@
-{"version":3,"names":["NativeModules","Platform","LINKING_ERROR","select","ios","default","RNWhisper","Proxy","get","Error","WhisperContext","constructor","id","transcribe","path","options","arguments","length","undefined","then","result","release","releaseContext","initWhisper","filePath","initContext","releaseAllWhisper","releaseAllContexts"],"sourceRoot":"../../src","sources":["index.tsx"],"mappings":"AAAA,SAASA,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AAEtD,MAAMC,aAAa,GAChB,sEAAqED,QAAQ,CAACE,MAAM,CAAC;EAAEC,GAAG,EAAE,gCAAgC;EAAEC,OAAO,EAAE;AAAG,CAAC,CAC3I,oDAAmD;AAEtD,MAAMC,SAAS,GAAGN,aAAa,CAACM,SAAS,GACrCN,aAAa,CAACM,SAAS,GACvB,IAAIC,KAAK,CACT,CAAC,CAAC,EACF;EACEC,GAAGA,CAAA,EAAG;IACJ,MAAM,IAAIC,KAAK,CAACP,aAAa,CAAC;EAChC;AACF,CAAC,CACF;AAoBH,MAAMQ,cAAc,CAAC;EAGnBC,WAAWA,CAACC,EAAU,EAAE;IACtB,IAAI,CAACA,EAAE,GAAGA,EAAE;EACd;EAEA,MAAMC,UAAUA,CAACC,IAAY,EAA8D;IAAA,IAA5DC,OAA0B,GAAAC,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;IAC5D,OAAOV,SAAS,CAACO,UAAU,CAAC,IAAI,CAACD,EAAE,EAAEE,IAAI,EAAEC,OAAO,CAAC,CAACI,IAAI,CAAEC,MAAc,KAAM;MAC5EA;IACF,CAAC,CAAC,CAAC;EACL;EAEA,MAAMC,OAAOA,CAAA,EAAG;IACd,OAAOf,SAAS,CAACgB,cAAc,CAAC,IAAI,CAACV,EAAE,CAAC;EAC1C;AACF;AAEA,OAAO,eAAeW,WAAWA,CAAA,EAEN;EAAA,IADzB;IAAEC;EAAgC,CAAC,GAAAR,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;EAExC,MAAMJ,EAAE,GAAG,MAAMN,SAAS,CAACmB,WAAW,CAACD,QAAQ,CAAC;EAChD,OAAO,IAAId,cAAc,CAACE,EAAE,CAAC;AAC/B;AAEA,OAAO,eAAec,iBAAiBA,CAAA,EAAkB;EACvD,OAAOpB,SAAS,CAACqB,kBAAkB,EAAE;AACvC"}
+{"version":3,"names":["NativeModules","Platform","LINKING_ERROR","select","ios","default","RNWhisper","Proxy","get","Error","WhisperContext","constructor","id","transcribe","path","options","arguments","length","undefined","release","releaseContext","initWhisper","filePath","initContext","releaseAllWhisper","releaseAllContexts"],"sourceRoot":"../../src","sources":["index.tsx"],"mappings":"AAAA,SAASA,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AAEtD,MAAMC,aAAa,GAChB,sEAAqED,QAAQ,CAACE,MAAM,CAAC;EAAEC,GAAG,EAAE,gCAAgC;EAAEC,OAAO,EAAE;AAAG,CAAC,CAC3I,oDAAmD;AAEtD,MAAMC,SAAS,GAAGN,aAAa,CAACM,SAAS,GACrCN,aAAa,CAACM,SAAS,GACvB,IAAIC,KAAK,CACT,CAAC,CAAC,EACF;EACEC,GAAGA,CAAA,EAAG;IACJ,MAAM,IAAIC,KAAK,CAACP,aAAa,CAAC;EAChC;AACF,CAAC,CACF;AA6BH,MAAMQ,cAAc,CAAC;EAGnBC,WAAWA,CAACC,EAAU,EAAE;IACtB,IAAI,CAACA,EAAE,GAAGA,EAAE;EACd;EAEA,MAAMC,UAAUA,CAACC,IAAY,EAA8D;IAAA,IAA5DC,OAA0B,GAAAC,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;IAC5D,OAAOV,SAAS,CAACO,UAAU,CAAC,IAAI,CAACD,EAAE,EAAEE,IAAI,EAAEC,OAAO,CAAC;EACrD;EAEA,MAAMI,OAAOA,CAAA,EAAG;IACd,OAAOb,SAAS,CAACc,cAAc,CAAC,IAAI,CAACR,EAAE,CAAC;EAC1C;AACF;AAEA,OAAO,eAAeS,WAAWA,CAAA,EAEN;EAAA,IADzB;IAAEC;EAAgC,CAAC,GAAAN,SAAA,CAAAC,MAAA,QAAAD,SAAA,QAAAE,SAAA,GAAAF,SAAA,MAAG,CAAC,CAAC;EAExC,MAAMJ,EAAE,GAAG,MAAMN,SAAS,CAACiB,WAAW,CAACD,QAAQ,CAAC;EAChD,OAAO,IAAIZ,cAAc,CAACE,EAAE,CAAC;AAC/B;AAEA,OAAO,eAAeY,iBAAiBA,CAAA,EAAkB;EACvD,OAAOlB,SAAS,CAACmB,kBAAkB,EAAE;AACvC"}

package/lib/typescript/index.d.ts

@@ -1,7 +1,10 @@
 export type TranscribeOptions = {
+    language?: string;
+    translate?: boolean;
     maxThreads?: number;
     maxContext?: number;
     maxLen?: number;
+    tokenTimestamps?: boolean;
     offset?: number;
     duration?: number;
     wordThold?: number;
@@ -10,9 +13,15 @@ export type TranscribeOptions = {
     beamSize?: number;
     bestOf?: number;
     speedUp?: boolean;
+    prompt?: string;
 };
 export type TranscribeResult = {
     result: string;
+    segments: Array<{
+        text: string;
+        t0: number;
+        t1: number;
+    }>;
 };
 declare class WhisperContext {
     id: number;

package/lib/typescript/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/index.tsx"],"names":[],"mappings":"AAiBA,MAAM,MAAM,iBAAiB,GAAG;IAC9B,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,OAAO,CAAC;CACnB,CAAA;AAED,MAAM,MAAM,gBAAgB,GAAG;IAC7B,MAAM,EAAE,MAAM,CAAC;CAChB,CAAA;AAED,cAAM,cAAc;IAClB,EAAE,EAAE,MAAM,CAAA;gBAEE,EAAE,EAAE,MAAM;IAIhB,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,GAAE,iBAAsB,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAMpF,OAAO;CAGd;AAED,wBAAsB,WAAW,CAC/B,EAAE,QAAQ,EAAE,GAAE;IAAE,QAAQ,CAAC,EAAE,MAAM,CAAA;CAAO,GACvC,OAAO,CAAC,cAAc,CAAC,CAGzB;AAED,wBAAsB,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC,CAEvD"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/index.tsx"],"names":[],"mappings":"AAiBA,MAAM,MAAM,iBAAiB,GAAG;IAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB,CAAA;AAED,MAAM,MAAM,gBAAgB,GAAG;IAC7B,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,KAAK,CAAC;QACd,IAAI,EAAE,MAAM,CAAC;QACb,EAAE,EAAE,MAAM,CAAC;QACX,EAAE,EAAE,MAAM,CAAC;KACZ,CAAC,CAAC;CACJ,CAAA;AAED,cAAM,cAAc;IAClB,EAAE,EAAE,MAAM,CAAA;gBAEE,EAAE,EAAE,MAAM;IAIhB,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,GAAE,iBAAsB,GAAG,OAAO,CAAC,gBAAgB,CAAC;IAIpF,OAAO;CAGd;AAED,wBAAsB,WAAW,CAC/B,EAAE,QAAQ,EAAE,GAAE;IAAE,QAAQ,CAAC,EAAE,MAAM,CAAA;CAAO,GACvC,OAAO,CAAC,cAAc,CAAC,CAGzB;AAED,wBAAsB,iBAAiB,IAAI,OAAO,CAAC,IAAI,CAAC,CAEvD"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "whisper.rn",
-  "version": "0.1.3",
+  "version": "0.1.4",
   "description": "React Native binding of whisper.cpp",
   "main": "lib/commonjs/index",
   "module": "lib/module/index",
@@ -98,13 +98,17 @@
       "@commitlint/config-conventional"
     ]
   },
+  "publishConfig": {
+    "registry": "https://registry.npmjs.org/"
+  },
   "release-it": {
     "git": {
       "commitMessage": "chore: release ${version}",
       "tagName": "v${version}"
     },
     "npm": {
-      "publish": true
+      "publish": true,
+      "skipChecks": true
     },
     "github": {
       "release": true

package/src/index.tsx

@@ -16,9 +16,12 @@ const RNWhisper = NativeModules.RNWhisper
   )
 
 export type TranscribeOptions = {
+  language?: string,
+  translate?: boolean,
   maxThreads?: number,
   maxContext?: number,
   maxLen?: number,
+  tokenTimestamps?: boolean,
   offset?: number,
   duration?: number,
   wordThold?: number,
@@ -27,10 +30,16 @@ export type TranscribeOptions = {
   beamSize?: number,
   bestOf?: number,
   speedUp?: boolean,
+  prompt?: string,
 }
 
 export type TranscribeResult = {
   result: string,
+  segments: Array<{
+    text: string,
+    t0: number,
+    t1: number,
+  }>,
 }
 
 class WhisperContext {
@@ -41,9 +50,7 @@ class WhisperContext {
   }
 
   async transcribe(path: string, options: TranscribeOptions = {}): Promise<TranscribeResult> {
-    return RNWhisper.transcribe(this.id, path, options).then((result: string) => ({
-      result
-    }))
+    return RNWhisper.transcribe(this.id, path, options)
   }
 
   async release() {