split audio and processing it step by step

app/audio_processor.py

@@ -95,7 +95,7 @@ class StuckAudioResender:
     Class to resend stuck raw audio files.
     """
 
-    def __init__(self, resend_stuck_audio_timeout_seconds=300, is_stuck_predicate=default_is_stuck_predicate):
+    def __init__(self, resend_stuck_audio_timeout_seconds=30, is_stuck_predicate=default_is_stuck_predicate):
         self._resend_stuck_audio_timeout_seconds = resend_stuck_audio_timeout_seconds
         self._is_stuck_predicate = is_stuck_predicate
 

app/audio_recognizer.py

@@ -1,6 +1,8 @@
 import asyncio
 import json
 import wave
+from pydub import AudioSegment
+from io import BytesIO
 
 import requests
 
@@ -44,7 +46,22 @@ def recognize(self, audio):
         recognized_words = list(map(self.parse_recognizer_result, recognizer_results))
         return RecognizedAudio(recognized_words)
 
-    def send_audio_to_recognizer(self, audio, language='ru'):
+    def send_audio_to_recognizer(self, audio_file, language='ru'):
+        audio_data = audio_file.read()
+        audio_file.close()
+
+        audio = AudioSegment.from_file(BytesIO(audio_data), format="mp3")
+        duration_seconds = audio.duration_seconds
+
+        segments = []
+        start_time = 0
+        while start_time < duration_seconds:
+            end_time = min(start_time + 10, duration_seconds)
+            segment = audio[start_time * 1000: end_time * 1000]
+            segments.append((segment, start_time))
+            start_time = end_time
+
+        # Параметры запроса
         params = {
             'task': 'transcribe',
             'language': language,
@@ -53,18 +70,19 @@ def send_audio_to_recognizer(self, audio, language='ru'):
         }
         headers = {'accept': 'application/json'}
 
-        audio_to_recognize_buffer = audio.read()
-        audio.close()
-
-        try:
-            files = {'audio_file': ("student_speech", audio_to_recognize_buffer, 'audio/mpeg')}
-            response = requests.post(self._url, params=params, headers=headers, files=files)
-            response.raise_for_status()
-        except requests.exceptions.RequestException as e:
-            logger.info(f"Recognition error occurred while processing audio file: {e}")
-            return []
-
-        data = response.json()
+        # Распознавание речи по сегментам
+        recognizer_results = []
+        for segment, segment_start_time in segments:
+            audio_to_recognize_buffer = segment.export(format="mp3").read()
+            try:
+                files = {'audio_file': ("student_speech", audio_to_recognize_buffer, 'audio/mpeg')}
+                response = requests.post(self._url, params=params, headers=headers, files=files)
+                response.raise_for_status()
+            except requests.exceptions.RequestException as e:
+                logger.info(f"Recognition error occurred while processing audio file: {e}")
+                return []
+
+            data = response.json()
 
         recognizer_results = []
         for segment in data["segments"]: