Preserve VibeVoice ASR turn text

Author: 0xShug0

app/server/runtime.cpp

@@ -379,7 +379,11 @@ std::string task_result_json_with_timing(
             const auto & segment = result.speech_segments[i];
             out << "{\"start_sample\":" << segment.span.start_sample
                 << ",\"end_sample\":" << segment.span.end_sample
-                << ",\"confidence\":" << segment.confidence << "}";
+                << ",\"confidence\":" << segment.confidence;
+            if (!segment.text.empty()) {
+                out << ",\"text\":" << json_quote(segment.text);
+            }
+            out << "}";
         }
         out << "]";
     }
@@ -394,7 +398,11 @@ std::string task_result_json_with_timing(
             out << "{\"start_sample\":" << turn.span.start_sample
                 << ",\"end_sample\":" << turn.span.end_sample
                 << ",\"speaker_id\":" << json_quote(turn.speaker_id)
-                << ",\"confidence\":" << turn.confidence << "}";
+                << ",\"confidence\":" << turn.confidence;
+            if (!turn.text.empty()) {
+                out << ",\"text\":" << json_quote(turn.text);
+            }
+            out << "}";
         }
         out << "]";
     }

app/workflow/file_sink.cpp

@@ -35,8 +35,11 @@ std::string speech_segments_to_json(const std::vector<engine::runtime::SpeechSeg
         }
         out << "{\"start_sample\":" << segments[i].span.start_sample
             << ",\"end_sample\":" << segments[i].span.end_sample
-            << ",\"confidence\":" << segments[i].confidence
-            << "}";
+            << ",\"confidence\":" << segments[i].confidence;
+        if (!segments[i].text.empty()) {
+            out << ",\"text\":" << quote_json(segments[i].text);
+        }
+        out << "}";
     }
     out << "]";
     return out.str();
@@ -52,8 +55,11 @@ std::string speaker_turns_to_json(const std::vector<engine::runtime::SpeakerTurn
         out << "{\"start_sample\":" << turns[i].span.start_sample
             << ",\"end_sample\":" << turns[i].span.end_sample
             << ",\"speaker_id\":" << quote_json(turns[i].speaker_id)
-            << ",\"confidence\":" << turns[i].confidence
-            << "}";
+            << ",\"confidence\":" << turns[i].confidence;
+        if (!turns[i].text.empty()) {
+            out << ",\"text\":" << quote_json(turns[i].text);
+        }
+        out << "}";
     }
     out << "]";
     return out.str();

assets/resources/four_speaker_short.txt

@@ -0,0 +1,4 @@
+Speaker 1: Welcome back to the studio. The agenda today is simple: check the solar forecast, test the backup radios, and agree on tomorrow's route.
+Speaker 2: I reviewed the forecast this morning. The east ridge should stay clear, but the valley may get fog after sunrise.
+Speaker 3: The radios are charged and paired. I also labeled the spare batteries so nobody mixes fresh cells with used ones.
+Speaker 4: Good. I will update the route map, mark the fog zone, and send everyone the final plan before dinner.

assets/resources/four_speaker_short.wav

@@ -89,12 +89,14 @@ struct TimeSpan {
 struct SpeechSegment {
     TimeSpan span;
     float confidence = 0.0f;
+    std::string text;
 };
 
 struct SpeakerTurn {
     TimeSpan span;
     std::string speaker_id;
     float confidence = 0.0f;
+    std::string text;
 };
 
 struct WordTimestamp {

include/engine/framework/runtime/session.h

@@ -547,6 +547,7 @@ runtime::TaskResult SileroRuntime::build_offline_result(
         result.speech_segments.push_back({
             {segment.start, segment.end},
             1.0f,
+            {},
         });
     }
     return result;
@@ -601,7 +602,7 @@ runtime::StreamEvent SileroRuntime::build_stream_event(
             runtime::VoiceActivityEvent::Kind::SpeechEnd,
             speech_end,
             probability,
-            runtime::SpeechSegment{{segment.start, segment.end}, probability},
+            runtime::SpeechSegment{{segment.start, segment.end}, probability, {}},
         });
     }
     return event;
@@ -649,13 +650,15 @@ runtime::TaskResult SileroRuntime::finalize_stream(const SileroVADConfig & confi
         result.speech_segments.push_back({
             {segment.start, segment.end},
             config.threshold,
+            {},
         });
     }
     const int64_t min_speech_samples = static_cast<int64_t>(kSampleRate) * config.min_speech_duration_ms / 1000;
     if (triggered_ && (current_sample_ - current_speech_start_) > min_speech_samples) {
         result.speech_segments.push_back({
             {current_speech_start_, current_sample_},
             config.threshold,
+            {},
         });
     }
     engine::debug::timing_log_scalar("silero_vad.streaming.finalize_ms", engine::debug::elapsed_ms(total_start));

src/models/silero_vad/runtime.cpp

@@ -968,10 +968,12 @@ runtime::TaskResult VibeVoiceASRSession::run_single(const VibeVoiceASRRequest &
         runtime::SpeechSegment speech_segment;
         speech_segment.span.start_sample = static_cast<int64_t>(segment.start_time * audio.sample_rate);
         speech_segment.span.end_sample = static_cast<int64_t>(segment.end_time * audio.sample_rate);
+        speech_segment.text = segment.text;
         result.speech_segments.push_back(speech_segment);
         runtime::SpeakerTurn turn;
         turn.span = speech_segment.span;
         turn.speaker_id = segment.speaker_id;
+        turn.text = segment.text;
         result.speaker_turns.push_back(std::move(turn));
     }