translator но без озвучивания слов на английском
This commit is contained in:
62
stt.py
62
stt.py
@@ -3,6 +3,7 @@ Speech-to-Text module using Deepgram API.
|
||||
Recognizes speech from microphone using streaming WebSocket.
|
||||
Supports Russian (default) and English.
|
||||
"""
|
||||
|
||||
import os
|
||||
import asyncio
|
||||
import threading
|
||||
@@ -20,6 +21,7 @@ from deepgram import (
|
||||
# Configure logging to suppress debug noise
|
||||
logging.getLogger("deepgram").setLevel(logging.WARNING)
|
||||
|
||||
|
||||
class SpeechRecognizer:
|
||||
"""Speech recognizer using Deepgram streaming."""
|
||||
|
||||
@@ -29,18 +31,18 @@ class SpeechRecognizer:
|
||||
self.stream = None
|
||||
self.transcript = ""
|
||||
self.lock = threading.Lock()
|
||||
|
||||
|
||||
def initialize(self):
|
||||
"""Initialize Deepgram client and PyAudio."""
|
||||
if not DEEPGRAM_API_KEY:
|
||||
raise ValueError("DEEPGRAM_API_KEY is not set in environment or config.")
|
||||
|
||||
|
||||
print("📦 Инициализация Deepgram STT...")
|
||||
config = DeepgramClientOptions(
|
||||
verbose=logging.WARNING,
|
||||
)
|
||||
self.dg_client = DeepgramClient(DEEPGRAM_API_KEY, config)
|
||||
|
||||
|
||||
self.pa = pyaudio.PyAudio()
|
||||
print("✅ Deepgram клиент готов")
|
||||
|
||||
@@ -59,13 +61,14 @@ class SpeechRecognizer:
|
||||
async def _process_audio(self, dg_connection, timeout_seconds, detection_timeout):
|
||||
"""Async loop to send audio and wait for results."""
|
||||
self.transcript = ""
|
||||
|
||||
transcript_parts = []
|
||||
|
||||
loop = asyncio.get_running_loop()
|
||||
stream = self._get_stream()
|
||||
|
||||
|
||||
stop_event = asyncio.Event()
|
||||
speech_started_event = asyncio.Event()
|
||||
|
||||
|
||||
# We need access to the outer 'self' (SpeechRecognizer instance)
|
||||
speech_recognizer_self = self
|
||||
|
||||
@@ -74,9 +77,11 @@ class SpeechRecognizer:
|
||||
if len(sentence) == 0:
|
||||
return
|
||||
if result.is_final:
|
||||
print(f"📝 Частичный результат: {sentence}")
|
||||
with speech_recognizer_self.lock:
|
||||
speech_recognizer_self.transcript = sentence
|
||||
transcript_parts.append(sentence)
|
||||
speech_recognizer_self.transcript = " ".join(
|
||||
transcript_parts
|
||||
).strip()
|
||||
|
||||
def on_speech_started(unused_self, speech_started, **kwargs):
|
||||
loop.call_soon_threadsafe(speech_started_event.set)
|
||||
@@ -102,7 +107,7 @@ class SpeechRecognizer:
|
||||
channels=1,
|
||||
sample_rate=SAMPLE_RATE,
|
||||
interim_results=True,
|
||||
utterance_end_ms="1200",
|
||||
utterance_end_ms=1200,
|
||||
vad_events=True,
|
||||
)
|
||||
|
||||
@@ -133,38 +138,45 @@ class SpeechRecognizer:
|
||||
print(f"\n🛑 Stream stopped. Chunks sent: {chunks_sent}")
|
||||
|
||||
sender_task = asyncio.create_task(send_audio())
|
||||
|
||||
|
||||
try:
|
||||
# 1. Wait for speech to start (detection_timeout)
|
||||
if detection_timeout:
|
||||
try:
|
||||
await asyncio.wait_for(speech_started_event.wait(), timeout=detection_timeout)
|
||||
await asyncio.wait_for(
|
||||
speech_started_event.wait(), timeout=detection_timeout
|
||||
)
|
||||
except asyncio.TimeoutError:
|
||||
# print("Detection timeout - no speech")
|
||||
stop_event.set()
|
||||
|
||||
|
||||
# 2. If started (or no detection timeout), wait for completion
|
||||
if not stop_event.is_set():
|
||||
await asyncio.wait_for(stop_event.wait(), timeout=timeout_seconds)
|
||||
|
||||
|
||||
except asyncio.TimeoutError:
|
||||
# print("Global timeout")
|
||||
pass
|
||||
|
||||
|
||||
stop_event.set()
|
||||
await sender_task
|
||||
# Finish is synchronous
|
||||
dg_connection.finish()
|
||||
|
||||
|
||||
return self.transcript
|
||||
|
||||
def listen(self, timeout_seconds: float = 7.0, detection_timeout: float = None, lang: str = "ru") -> str:
|
||||
def listen(
|
||||
self,
|
||||
timeout_seconds: float = 7.0,
|
||||
detection_timeout: float = None,
|
||||
lang: str = "ru",
|
||||
) -> str:
|
||||
"""
|
||||
Listen to microphone and transcribe speech.
|
||||
"""
|
||||
if not self.dg_client:
|
||||
self.initialize()
|
||||
|
||||
|
||||
self.current_lang = lang
|
||||
print(f"🎙️ Слушаю ({lang})...")
|
||||
|
||||
@@ -172,16 +184,18 @@ class SpeechRecognizer:
|
||||
dg_connection = self.dg_client.listen.live.v("1")
|
||||
|
||||
try:
|
||||
transcript = asyncio.run(self._process_audio(dg_connection, timeout_seconds, detection_timeout))
|
||||
|
||||
transcript = asyncio.run(
|
||||
self._process_audio(dg_connection, timeout_seconds, detection_timeout)
|
||||
)
|
||||
|
||||
final_text = transcript.strip() if transcript else ""
|
||||
if final_text:
|
||||
print(f"📝 Распознано: {final_text}")
|
||||
else:
|
||||
print("⚠️ Речь не распознана")
|
||||
|
||||
|
||||
return final_text
|
||||
|
||||
|
||||
except Exception as e:
|
||||
print(f"❌ Ошибка STT: {e}")
|
||||
return ""
|
||||
@@ -208,7 +222,9 @@ def get_recognizer() -> SpeechRecognizer:
|
||||
return _recognizer
|
||||
|
||||
|
||||
def listen(timeout_seconds: float = 7.0, detection_timeout: float = None, lang: str = "ru") -> str:
|
||||
def listen(
|
||||
timeout_seconds: float = 7.0, detection_timeout: float = None, lang: str = "ru"
|
||||
) -> str:
|
||||
"""Listen to microphone and return transcribed text."""
|
||||
return get_recognizer().listen(timeout_seconds, detection_timeout, lang)
|
||||
|
||||
@@ -218,4 +234,4 @@ def cleanup():
|
||||
global _recognizer
|
||||
if _recognizer:
|
||||
_recognizer.cleanup()
|
||||
_recognizer = None
|
||||
_recognizer = None
|
||||
|
||||
Reference in New Issue
Block a user