flare / stt /stt_google.py
ciyidogan's picture
Update stt/stt_google.py
6abf273 verified
raw
history blame
10.2 kB
"""
Google Cloud Speech-to-Text Implementation - Simple Batch Mode
"""
from typing import Optional, List
from datetime import datetime
import io
import wave
from google.cloud import speech
from google.cloud.speech import RecognitionConfig, RecognitionAudio
from utils.logger import log_info, log_error, log_debug, log_warning
from .stt_interface import STTInterface, STTConfig, TranscriptionResult
class GoogleSTT(STTInterface):
def __init__(self, credentials_path: Optional[str] = None):
"""
Initialize Google STT
Args:
credentials_path: Path to service account JSON file (optional if using default credentials)
"""
try:
# Initialize client
if credentials_path:
self.client = speech.SpeechClient.from_service_account_file(credentials_path)
log_info(f"✅ Google STT initialized with service account: {credentials_path}")
else:
# Use default credentials (ADC)
self.client = speech.SpeechClient()
log_info("✅ Google STT initialized with default credentials")
except Exception as e:
log_error(f"❌ Failed to initialize Google STT: {str(e)}")
raise
def _map_language_code(self, language: str) -> str:
"""Map language codes to Google format"""
# Google uses BCP-47 language codes
language_map = {
"tr": "tr-TR",
"tr-TR": "tr-TR",
"en": "en-US",
"en-US": "en-US",
"en-GB": "en-GB",
"de": "de-DE",
"de-DE": "de-DE",
"fr": "fr-FR",
"fr-FR": "fr-FR",
"es": "es-ES",
"es-ES": "es-ES",
"it": "it-IT",
"it-IT": "it-IT",
"pt": "pt-BR",
"pt-BR": "pt-BR",
"ru": "ru-RU",
"ru-RU": "ru-RU",
"ja": "ja-JP",
"ja-JP": "ja-JP",
"ko": "ko-KR",
"ko-KR": "ko-KR",
"zh": "zh-CN",
"zh-CN": "zh-CN",
"ar": "ar-SA",
"ar-SA": "ar-SA",
}
# Default to the language itself if not in map
return language_map.get(language, language)
async def transcribe(self, audio_data: bytes, config: STTConfig) -> Optional[TranscriptionResult]:
try:
if not audio_data:
log_warning("⚠️ No audio data provided")
return None
log_info(f"📊 Transcribing {len(audio_data)} bytes of audio")
# ✅ Detaylı audio analizi - logda
import struct
samples = struct.unpack(f'{len(audio_data)//2}h', audio_data)
total_samples = len(samples)
# 1. Genel istatistikler
non_zero_samples = [s for s in samples if s != 0]
zero_count = total_samples - len(non_zero_samples)
if non_zero_samples:
avg_amplitude = sum(abs(s) for s in non_zero_samples) / len(non_zero_samples)
max_amplitude = max(abs(s) for s in non_zero_samples)
else:
avg_amplitude = 0
max_amplitude = 0
log_info(f"🔍 Audio stats: {total_samples} total samples, {zero_count} zeros ({zero_count/total_samples:.1%})")
log_info(f"🔍 Non-zero stats: avg={avg_amplitude:.1f}, max={max_amplitude}")
# 2. Bölüm bazlı analiz (10 bölüme ayır)
section_size = total_samples // 10
log_info(f"🔍 Section analysis (each {section_size} samples):")
for i in range(10):
start_idx = i * section_size
end_idx = (i + 1) * section_size if i < 9 else total_samples
section = samples[start_idx:end_idx]
section_non_zero = [s for s in section if s != 0]
section_max = max(abs(s) for s in section_non_zero) if section_non_zero else 0
section_avg = sum(abs(s) for s in section_non_zero) / len(section_non_zero) if section_non_zero else 0
zero_ratio = (len(section) - len(section_non_zero)) / len(section)
log_info(f" Section {i+1}: max={section_max}, avg={section_avg:.1f}, zeros={zero_ratio:.1%}")
# 3. İlk konuşma başlangıcını bul
speech_threshold = 500 # RMS eşiği
speech_start_idx = -1
# 100 sample'lık pencerelerle RMS hesapla
window_size = 100
for i in range(0, total_samples - window_size, window_size):
window = samples[i:i + window_size]
rms = (sum(s * s for s in window) / window_size) ** 0.5
if rms > speech_threshold:
speech_start_idx = i
break
if speech_start_idx >= 0:
speech_start_time = speech_start_idx / config.sample_rate
log_info(f"🎤 Speech detected starting at sample {speech_start_idx} ({speech_start_time:.2f}s)")
else:
log_warning("⚠️ No speech detected above threshold in entire audio")
# 4. Audio'nun gerçekten boş olup olmadığını kontrol et
if max_amplitude < 100:
log_warning(f"⚠️ Audio appears silent: max_amplitude={max_amplitude}")
return None
if zero_count / total_samples > 0.95: # %95'den fazla sıfır
log_warning(f"⚠️ Audio is mostly zeros: {zero_count/total_samples:.1%}")
return None
# Convert to WAV format
wav_audio = self._convert_to_wav(audio_data, config.sample_rate)
# Configure recognition
recognition_config = RecognitionConfig(
encoding=RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=16000,
language_code="tr-TR",
audio_channel_count=1,
enable_separate_recognition_per_channel=False,
enable_automatic_punctuation=True,
)
# Create audio object
audio = RecognitionAudio(content=wav_audio)
# Perform synchronous recognition
log_info(f"🔄 Sending audio to Google Cloud Speech API...")
response = self.client.recognize(config=recognition_config, audio=audio)
# ✅ Detaylı response analizi
log_info(f"🔍 Google response details:")
log_info(f" - Has results: {bool(response.results)}")
log_info(f" - Results count: {len(response.results) if response.results else 0}")
if hasattr(response, 'total_billed_time'):
if response.total_billed_time and response.total_billed_time.total_seconds() > 0:
log_info(f" - Billed time: {response.total_billed_time.total_seconds()}s")
else:
log_info(f" - Billed time: 0s (no audio processed)")
# Process results
if response.results and len(response.results) > 0:
for i, result in enumerate(response.results):
log_info(f" - Result {i}: {len(result.alternatives)} alternatives")
if result.alternatives:
for j, alt in enumerate(result.alternatives):
log_info(f" - Alt {j}: '{alt.transcript}' (conf: {alt.confidence:.3f})")
result = response.results[0]
if result.alternatives and len(result.alternatives) > 0:
alternative = result.alternatives[0]
transcription = TranscriptionResult(
text=alternative.transcript,
confidence=alternative.confidence,
timestamp=datetime.now().timestamp(),
language="tr-TR",
word_timestamps=None
)
log_info(f"✅ Transcription SUCCESS: '{alternative.transcript}' (confidence: {alternative.confidence:.2f})")
return transcription
log_warning("⚠️ No transcription results - Google couldn't recognize speech")
return None
except Exception as e:
log_error(f"❌ Error during transcription: {str(e)}")
import traceback
log_error(f"Traceback: {traceback.format_exc()}")
return None
def _convert_to_wav(self, audio_data: bytes, sample_rate: int) -> bytes:
"""Convert raw PCM audio to WAV format"""
# Create WAV file in memory
wav_buffer = io.BytesIO()
with wave.open(wav_buffer, 'wb') as wav_file:
# Set WAV parameters
wav_file.setnchannels(1) # Mono
wav_file.setsampwidth(2) # 16-bit
wav_file.setframerate(sample_rate)
wav_file.writeframes(audio_data)
# Get WAV data
wav_buffer.seek(0)
return wav_buffer.read()
def get_supported_languages(self) -> List[str]:
"""Get list of supported language codes"""
# Google Cloud Speech-to-Text supported languages (partial list)
return [
"tr-TR", "en-US", "en-GB", "en-AU", "en-CA", "en-IN",
"es-ES", "es-MX", "es-AR", "fr-FR", "fr-CA", "de-DE",
"it-IT", "pt-BR", "pt-PT", "ru-RU", "ja-JP", "ko-KR",
"zh-CN", "zh-TW", "ar-SA", "ar-EG", "hi-IN", "nl-NL",
"pl-PL", "sv-SE", "da-DK", "no-NO", "fi-FI", "el-GR",
"he-IL", "th-TH", "vi-VN", "id-ID", "ms-MY", "fil-PH"
]
def get_provider_name(self) -> str:
"""Get provider name"""
return "google"