output segments

Author: litongmacos

stream_components_params.h

@@ -18,7 +18,7 @@ namespace stream_components {
     int32_t n_samples_keep = 0;
     int32_t n_samples_len = 0;
     int32_t n_samples_30s = 0;
-    bool use_vad = false;
+    bool use_vad = true;
 
     float vad_thold = 0.6f;
     float freq_thold = 100.0f;

stream_components_service.cpp

@@ -41,7 +41,7 @@ WhisperService::~WhisperService() {
   whisper_free(ctx);
 }
 
-WhisperOutputPtr WhisperService::process(const float *samples, int sample_count) {
+bool WhisperService::process(const float *samples, int sample_count) {
   whisper_full_params wparams = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
 
   wparams.print_progress = false;
@@ -70,7 +70,7 @@ WhisperOutputPtr WhisperService::process(const float *samples, int sample_count)
 
   // *** Run the actual inference!!! ***
   if (whisper_full(ctx, wparams, samples, sample_count) != 0) {
-    return nullptr;
+    return false;
   }
 
   // Now sure whether n_iter and n_new_line should have ever been there...
@@ -96,8 +96,5 @@ WhisperOutputPtr WhisperService::process(const float *samples, int sample_count)
       }
     }
   }
-
-  auto r = std::make_shared<WhisperStreamOutput>(ctx, server_params);
-
-  return r;
+  return true;
 }

stream_components_service.h

@@ -22,9 +22,7 @@ namespace stream_components {
 
     ~WhisperService();
 
-    WhisperOutputPtr process(
-      const float *samples,
-      int size);
+    bool process(const float *samples,int size);
 
     service_params server_params;
     audio_params audio_params;

stream_local.cpp

@@ -3,6 +3,7 @@
 #include "stream_components_params.h"
 #include "stream_components_output.h"
 #include "stream_components_service.h"
+//#include "json.hpp"
 
 using namespace stream_components;
 
@@ -101,6 +102,8 @@ int main(int argc, char **argv) {
 
   // Compute derived parameters
   params.initialize();
+  //output params
+  printf("vad:%d\n", params.audio.use_vad);
 
   // Check parameters
   if (params.service.language != "auto" && whisper_lang_id(params.service.language.c_str()) == -1) {
@@ -115,7 +118,7 @@ int main(int argc, char **argv) {
   // Instantiate the service
   struct whisper_context_params cparams;
   cparams.use_gpu = params.service.use_gpu;
-  stream_components::WhisperService whisperService(params.service, params.audio,cparams);
+  stream_components::WhisperService whisperService(params.service, params.audio, cparams);
 
   // Print the 'header'...
   WhisperStreamOutput::to_json(std::cout, params.service, whisperService.ctx);
@@ -133,12 +136,27 @@ int main(int argc, char **argv) {
     // get next microphone section
     auto pcmf32 = microphone.get_next();
 
+    // process
+    bool isOk = whisperService.process(pcmf32.data(), pcmf32.size());
+    printf("isOk:%d\n", isOk);
     // get the whisper output
-    const WhisperOutputPtr &outputPtr = whisperService.process(pcmf32.data(), pcmf32.size());
-    // write the output as json to stdout (for this example)
-    if (outputPtr) {
-      outputPtr->transcription_to_json(std::cout);
+    if (isOk) {
+//      WhisperOutputPtr outputPtr = std::make_shared<WhisperStreamOutput>(whisperService.ctx, params.service);
+//      // write the output as json to stdout (for this example)
+//      if (outputPtr) {
+//        outputPtr->transcription_to_json(std::cout);
+//      }
+      const int n_segments = whisper_full_n_segments(whisperService.ctx);
+      printf("n_segments:%d\n", n_segments);
+      for (int i = 0; i < n_segments; ++i) {
+        const char *text = whisper_full_get_segment_text(whisperService.ctx, i);
+        const int64_t t0 = whisper_full_get_segment_t0(whisperService.ctx, i);
+        const int64_t t1 = whisper_full_get_segment_t1(whisperService.ctx, i);
+        printf("%lld-->%lld:%s\n", t0, t1, text);
+      }
+
     }
+
   }
   std::cout << "EXITED MAIN LOOP" << std::endl;
   return 0;