whisper.rn 0.4.0-rc.5 → 0.4.0-rc.7

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

@@ -115,7 +115,7 @@ public class RNWhisper implements LifecycleEventListener {
         promise.resolve(result);
         tasks.remove(this);
       }
-    }.execute();
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
     tasks.put(task, "initContext");
   }
 
@@ -174,7 +174,7 @@ public class RNWhisper implements LifecycleEventListener {
         promise.resolve(data);
         tasks.remove(this);
       }
-    }.execute();
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
     tasks.put(task, "transcribeFile-" + id);
   }
 
@@ -231,7 +231,7 @@ public class RNWhisper implements LifecycleEventListener {
         promise.resolve(null);
         tasks.remove(this);
       }
-    }.execute();
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
     tasks.put(task, "abortTranscribe-" + id);
   }
 
@@ -272,7 +272,7 @@ public class RNWhisper implements LifecycleEventListener {
         promise.resolve(null);
         tasks.remove(this);
       }
-    }.execute();
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
     tasks.put(task, "releaseContext-" + id);
   }
 
@@ -299,7 +299,7 @@ public class RNWhisper implements LifecycleEventListener {
         promise.resolve(null);
         tasks.remove(this);
       }
-    }.execute();
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
     tasks.put(task, "releaseAllContexts");
   }
 

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

@@ -110,6 +110,9 @@ public class WhisperContext {
     final int audioSliceSec = realtimeAudioSliceSec > 0 && realtimeAudioSliceSec < audioSec ? realtimeAudioSliceSec : audioSec;
     isUseSlices = audioSliceSec < audioSec;
 
+    double realtimeAudioMinSec = options.hasKey("realtimeAudioMinSec") ? options.getDouble("realtimeAudioMinSec") : 0;
+    final double audioMinSec = realtimeAudioMinSec > 0.5 && realtimeAudioMinSec <= audioSliceSec ? realtimeAudioMinSec : 1;
+
     createRealtimeTranscribeJob(jobId, context, options);
 
     sliceNSamples = new ArrayList<Integer>();
@@ -144,7 +147,8 @@ public class WhisperContext {
                 ) {
                   finishRealtimeTranscribe(Arguments.createMap());
                 } else if (!isTranscribing) {
-                  if (!vad(sliceIndex, nSamples, 0)) {
+                  boolean isSamplesEnough = nSamples / SAMPLE_RATE >= audioMinSec;
+                  if (!isSamplesEnough || !vad(sliceIndex, nSamples, 0)) {
                     finishRealtimeTranscribe(Arguments.createMap());
                     break;
                   }
@@ -169,7 +173,8 @@ public class WhisperContext {
               nSamples += n;
               sliceNSamples.set(sliceIndex, nSamples);
 
-              if (!isSpeech) continue;
+              boolean isSamplesEnough = nSamples / SAMPLE_RATE >= audioMinSec;
+              if (!isSamplesEnough || !isSpeech) continue;
 
               if (!isTranscribing && nSamples > SAMPLE_RATE / 2) {
                 isTranscribing = true;

package/android/src/main/jni.cpp

@@ -199,7 +199,7 @@ struct whisper_full_params createFullParams(JNIEnv *env, jobject options) {
     params.print_progress = false;
     params.print_timestamps = false;
     params.print_special = false;
-    
+
     int max_threads = std::thread::hardware_concurrency();
     // Use 2 threads by default on 4-core devices, 4 threads on more cores
     int default_n_threads = max_threads == 4 ? 2 : min(4, max_threads);
@@ -307,7 +307,7 @@ Java_com_rnwhisper_WhisperContext_fullWithNewJob(
         // whisper_print_timings(context);
     }
     env->ReleaseFloatArrayElements(audio_data, audio_data_arr, JNI_ABORT);
-    
+
     if (job->is_aborted()) code = -999;
     rnwhisper::job_remove(job_id);
     return code;
@@ -339,6 +339,7 @@ Java_com_rnwhisper_WhisperContext_createRealtimeTranscribeJob(
         vad,
         readablemap::getInt(env, options, "realtimeAudioSec", 0),
         readablemap::getInt(env, options, "realtimeAudioSliceSec", 0),
+        readablemap::getFloat(env, options, "realtimeAudioMinSec", 0),
         audio_output_path_str
     );
 }

package/cpp/ggml-alloc.h

@@ -43,7 +43,7 @@ WSP_GGML_API size_t wsp_ggml_allocr_alloc_graph(wsp_ggml_allocr_t alloc, struct
 // ggml-backend v2 API
 //
 
-// Seperate tensor and graph allocator objects
+// Separate tensor and graph allocator objects
 // This is necessary for multi-backend allocation because the graph allocator needs to use multiple tensor allocators
 // The original API is kept as a wrapper around the new API