Add support for multiple languages for Azure transcriber (vocodedev#295)

* Add support for multiple languages for Azure transcriber * cleanup --------- Co-authored-by: Kian <kianhooshmand@berkeley.edu>

Author: applebaconsoda123

vocode/streaming/models/transcriber.py

@@ -13,6 +13,8 @@
 from .audio_encoding import AudioEncoding
 from .model import TypedModel
 
+AZURE_DEFAULT_LANGUAGE = "en-US"
+
 
 class TranscriberType(str, Enum):
  BASE = "transcriber_base"
@@ -119,7 +121,8 @@ class GoogleTranscriberConfig(TranscriberConfig, type=TranscriberType.GOOGLE.val
 
 
 class AzureTranscriberConfig(TranscriberConfig, type=TranscriberType.AZURE.value):
- pass
+ language: str = AZURE_DEFAULT_LANGUAGE
+ candidate_languages: Optional[List[str]] = None
 
 
 class AssemblyAITranscriberConfig(

vocode/streaming/transcriber/azure_transcriber.py

@@ -1,11 +1,8 @@
-import asyncio
 import logging
 import queue
 from typing import Optional
-import threading
 
 from azure.cognitiveservices.speech.audio import (
- AudioInputStream,
  PushAudioInputStream,
  AudioStreamFormat,
  AudioStreamWaveFormat,
@@ -54,9 +51,29 @@ def __init__(
  region=getenv("AZURE_SPEECH_REGION"),
  )
 
- self.speech = speechsdk.SpeechRecognizer(
- speech_config=speech_config, audio_config=config
- )
+ speech_params = {
+ "speech_config": speech_config,
+ "audio_config": config,
+ }
+
+ if self.transcriber_config.candidate_languages:
+ speech_config.set_property(
+ property_id=speechsdk.PropertyId.SpeechServiceConnection_LanguageIdMode,
+ value="Continuous",
+ )
+ auto_detect_source_language_config = (
+ speechsdk.languageconfig.AutoDetectSourceLanguageConfig(
+ languages=self.transcriber_config.candidate_languages
+ )
+ )
+
+ speech_params[
+ "auto_detect_source_language_config"
+ ] = auto_detect_source_language_config
+ else:
+ speech_params["language"] = self.transcriber_config.language
+
+ self.speech = speechsdk.SpeechRecognizer(**speech_params)
 
  self._ended = False
  self.is_ready = False