extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-04-18 16:28:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge 5c76377b09 into f92bd59951

Browse Source
This commit is contained in:
Ranjit 2025-03-30 06:08:41 +02:00 committed by GitHub
commit aaab74ea08
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 10 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
examples/whisper.android/app/src/main/java/com/whispercppdemo/ui/main/MainScreenViewModel.kt
View File

@ -143,7 +143,7 @@ class MainScreenViewModel(private val application: Application) : ViewModel() {
printMessage("${data.size / (16000 / 1000)} ms\n") printMessage("${data.size / (16000 / 1000)} ms\n")
printMessage("Transcribing data...\n") printMessage("Transcribing data...\n")
val start = System.currentTimeMillis() val start = System.currentTimeMillis()
val text = whisperContext?.transcribeData(data) val text = whisperContext?.transcribeData(data, tokenTimestamps = false, splitOnWord = false, maxLen = 1)
val elapsed = System.currentTimeMillis() - start val elapsed = System.currentTimeMillis() - start
printMessage("Done ($elapsed ms): \n$text\n") printMessage("Done ($elapsed ms): \n$text\n")
} catch (e: Exception) { } catch (e: Exception) {
@ -222,4 +222,4 @@ private suspend fun Context.copyData(
} }
Log.v(LOG_TAG, "Copied $assetPath to $destination") Log.v(LOG_TAG, "Copied $assetPath to $destination")
} }
} }

8
examples/whisper.android/lib/src/main/java/com/whispercpp/whisper/LibWhisper.kt
View File

@ -16,11 +16,11 @@ class WhisperContext private constructor(private var ptr: Long) {
Executors.newSingleThreadExecutor().asCoroutineDispatcher() Executors.newSingleThreadExecutor().asCoroutineDispatcher()
) )
suspend fun transcribeData(data: FloatArray, printTimestamp: Boolean = true): String = withContext(scope.coroutineContext) { suspend fun transcribeData(data: FloatArray, printTimestamp: Boolean = true, tokenTimestamps: Boolean = false, splitOnWord: Boolean = false, maxLen: Int): String = withContext(scope.coroutineContext) {
require(ptr != 0L) require(ptr != 0L)
val numThreads = WhisperCpuConfig.preferredThreadCount val numThreads = WhisperCpuConfig.preferredThreadCount
Log.d(LOG_TAG, "Selecting $numThreads threads") Log.d(LOG_TAG, "Selecting $numThreads threads")
WhisperLib.fullTranscribe(ptr, numThreads, data) WhisperLib.fullTranscribe(ptr, numThreads, data, tokenTimestamps, splitOnWord, maxLen)
val textCount = WhisperLib.getTextSegmentCount(ptr) val textCount = WhisperLib.getTextSegmentCount(ptr)
return@withContext buildString { return@withContext buildString {
for (i in 0 until textCount) { for (i in 0 until textCount) {
@ -134,7 +134,7 @@ private class WhisperLib {
external fun initContextFromAsset(assetManager: AssetManager, assetPath: String): Long external fun initContextFromAsset(assetManager: AssetManager, assetPath: String): Long
external fun initContext(modelPath: String): Long external fun initContext(modelPath: String): Long
external fun freeContext(contextPtr: Long) external fun freeContext(contextPtr: Long)
external fun fullTranscribe(contextPtr: Long, numThreads: Int, audioData: FloatArray) external fun fullTranscribe(contextPtr: Long, numThreads: Int, audioData: FloatArray, tokenTimestamps: Boolean, splitOnWord: Boolean, maxLen: Int)
external fun getTextSegmentCount(contextPtr: Long): Int external fun getTextSegmentCount(contextPtr: Long): Int
external fun getTextSegment(contextPtr: Long, index: Int): String external fun getTextSegment(contextPtr: Long, index: Int): String
external fun getTextSegmentT0(contextPtr: Long, index: Int): Long external fun getTextSegmentT0(contextPtr: Long, index: Int): Long
@ -177,4 +177,4 @@ private fun cpuInfo(): String? {
Log.w(LOG_TAG, "Couldn't read /proc/cpuinfo", e) Log.w(LOG_TAG, "Couldn't read /proc/cpuinfo", e)
null null
} }
} }

5
examples/whisper.android/lib/src/main/jni/whisper/jni.c
View File

@ -163,7 +163,7 @@ Java_com_whispercpp_whisper_WhisperLib_00024Companion_freeContext(
JNIEXPORT void JNICALL JNIEXPORT void JNICALL
Java_com_whispercpp_whisper_WhisperLib_00024Companion_fullTranscribe( Java_com_whispercpp_whisper_WhisperLib_00024Companion_fullTranscribe(
JNIEnv *env, jobject thiz, jlong context_ptr, jint num_threads, jfloatArray audio_data) { JNIEnv *env, jobject thiz, jlong context_ptr, jint num_threads, jfloatArray audio_data, jboolean token_timestamps, jboolean split_on_word, jint max_len) {
UNUSED(thiz); UNUSED(thiz);
struct whisper_context *context = (struct whisper_context *) context_ptr; struct whisper_context *context = (struct whisper_context *) context_ptr;
jfloat *audio_data_arr = (*env)->GetFloatArrayElements(env, audio_data, NULL); jfloat *audio_data_arr = (*env)->GetFloatArrayElements(env, audio_data, NULL);
@ -181,6 +181,9 @@ Java_com_whispercpp_whisper_WhisperLib_00024Companion_fullTranscribe(
params.offset_ms = 0; params.offset_ms = 0;
params.no_context = true; params.no_context = true;
params.single_segment = false; params.single_segment = false;
params.token_timestamps = token_timestamps;
params.split_on_word = split_on_word;
params.max_len = max_len;
whisper_reset_timings(context); whisper_reset_timings(context);