whisper.android : update example, add field to print timestamp (#2072)

ggerganov · May 13, 2024 · e93081f · e93081f
1 parent b6bbce4
commit e93081f
Show file tree

Hide file tree

Showing 3 changed files with 42 additions and 3 deletions.
diff --git a/examples/whisper.android/app/src/main/java/com/whispercppdemo/ui/main/MainScreenViewModel.kt b/examples/whisper.android/app/src/main/java/com/whispercppdemo/ui/main/MainScreenViewModel.kt
@@ -145,7 +145,7 @@ class MainScreenViewModel(private val application: Application) : ViewModel() {
             val start = System.currentTimeMillis()
             val text = whisperContext?.transcribeData(data)
             val elapsed = System.currentTimeMillis() - start
-            printMessage("Done ($elapsed ms): $text\n")
+            printMessage("Done ($elapsed ms): \n$text\n")
         } catch (e: Exception) {
             Log.w(LOG_TAG, e)
             printMessage("${e.localizedMessage}\n")

diff --git a/examples/whisper.android/lib/src/main/java/com/whispercpp/whisper/LibWhisper.kt b/examples/whisper.android/lib/src/main/java/com/whispercpp/whisper/LibWhisper.kt
@@ -16,15 +16,21 @@ class WhisperContext private constructor(private var ptr: Long) {
         Executors.newSingleThreadExecutor().asCoroutineDispatcher()
     )
 
-    suspend fun transcribeData(data: FloatArray): String = withContext(scope.coroutineContext) {
+    suspend fun transcribeData(data: FloatArray, printTimestamp: Boolean = true): String = withContext(scope.coroutineContext) {
         require(ptr != 0L)
         val numThreads = WhisperCpuConfig.preferredThreadCount
         Log.d(LOG_TAG, "Selecting $numThreads threads")
         WhisperLib.fullTranscribe(ptr, numThreads, data)
         val textCount = WhisperLib.getTextSegmentCount(ptr)
         return@withContext buildString {
             for (i in 0 until textCount) {
-                append(WhisperLib.getTextSegment(ptr, i))
+                if (printTimestamp) {
+                    val textTimestamp = "[${toTimestamp(WhisperLib.getTextSegmentT0(ptr, i))} --> ${toTimestamp(WhisperLib.getTextSegmentT1(ptr, i))}]"
+                    val textSegment = WhisperLib.getTextSegment(ptr, i)
+                    append("$textTimestamp: $textSegment\n")
+                } else {
+                    append(WhisperLib.getTextSegment(ptr, i))
+                }
             }
         }
     }
@@ -131,12 +137,29 @@ private class WhisperLib {
         external fun fullTranscribe(contextPtr: Long, numThreads: Int, audioData: FloatArray)
         external fun getTextSegmentCount(contextPtr: Long): Int
         external fun getTextSegment(contextPtr: Long, index: Int): String
+        external fun getTextSegmentT0(contextPtr: Long, index: Int): Long
+        external fun getTextSegmentT1(contextPtr: Long, index: Int): Long
         external fun getSystemInfo(): String
         external fun benchMemcpy(nthread: Int): String
         external fun benchGgmlMulMat(nthread: Int): String
     }
 }
 
+//  500 -> 00:05.000
+// 6000 -> 01:00.000
+private fun toTimestamp(t: Long, comma: Boolean = false): String {
+    var msec = t * 10
+    val hr = msec / (1000 * 60 * 60)
+    msec -= hr * (1000 * 60 * 60)
+    val min = msec / (1000 * 60)
+    msec -= min * (1000 * 60)
+    val sec = msec / 1000
+    msec -= sec * 1000
+
+    val delimiter = if (comma) "," else "."
+    return String.format("%02d:%02d:%02d%s%03d", hr, min, sec, delimiter, msec)
+}
+
 private fun isArmEabiV7a(): Boolean {
     return Build.SUPPORTED_ABIS[0].equals("armeabi-v7a")
 }

diff --git a/examples/whisper.android/lib/src/main/jni/whisper/jni.c b/examples/whisper.android/lib/src/main/jni/whisper/jni.c
@@ -212,6 +212,22 @@ Java_com_whispercpp_whisper_WhisperLib_00024Companion_getTextSegment(
     return string;
 }
 
+JNIEXPORT jlong JNICALL
+Java_com_whispercpp_whisper_WhisperLib_00024Companion_getTextSegmentT0(
+        JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
+    UNUSED(thiz);
+    struct whisper_context *context = (struct whisper_context *) context_ptr;
+    return whisper_full_get_segment_t0(context, index);
+}
+
+JNIEXPORT jlong JNICALL
+Java_com_whispercpp_whisper_WhisperLib_00024Companion_getTextSegmentT1(
+        JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
+    UNUSED(thiz);
+    struct whisper_context *context = (struct whisper_context *) context_ptr;
+    return whisper_full_get_segment_t1(context, index);
+}
+
 JNIEXPORT jstring JNICALL
 Java_com_whispercpp_whisper_WhisperLib_00024Companion_getSystemInfo(
         JNIEnv *env, jobject thiz