Merge pull request #1190 from pipecat-ai/mb/languages-hosted-whisper

Add language support to OpenAI and Groq hosted Whisper

src/pipecat/services/base_whisper.py

@@ -10,6 +10,7 @@ from loguru import logger
 
 from pipecat.frames.frames import ErrorFrame, Frame, TranscriptionFrame
 from pipecat.services.ai_services import SegmentedSTTService
+from pipecat.transcriptions.language import Language
 from pipecat.utils.time import time_now_iso8601
 
 try:
@@ -23,6 +24,82 @@ except ModuleNotFoundError as e:
     raise Exception(f"Missing module: {e}")
 
 
+def language_to_whisper_language(language: Language) -> Optional[str]:
+    """Language support for Whisper API.
+
+    Docs: https://platform.openai.com/docs/guides/speech-to-text#supported-languages
+    """
+    BASE_LANGUAGES = {
+        Language.AF: "af",
+        Language.AR: "ar",
+        Language.HY: "hy",
+        Language.AZ: "az",
+        Language.BE: "be",
+        Language.BS: "bs",
+        Language.BG: "bg",
+        Language.CA: "ca",
+        Language.ZH: "zh",
+        Language.HR: "hr",
+        Language.CS: "cs",
+        Language.DA: "da",
+        Language.NL: "nl",
+        Language.EN: "en",
+        Language.ET: "et",
+        Language.FI: "fi",
+        Language.FR: "fr",
+        Language.GL: "gl",
+        Language.DE: "de",
+        Language.EL: "el",
+        Language.HE: "he",
+        Language.HI: "hi",
+        Language.HU: "hu",
+        Language.IS: "is",
+        Language.ID: "id",
+        Language.IT: "it",
+        Language.JA: "ja",
+        Language.KN: "kn",
+        Language.KK: "kk",
+        Language.KO: "ko",
+        Language.LV: "lv",
+        Language.LT: "lt",
+        Language.MK: "mk",
+        Language.MS: "ms",
+        Language.MR: "mr",
+        Language.MI: "mi",
+        Language.NE: "ne",
+        Language.NO: "no",
+        Language.FA: "fa",
+        Language.PL: "pl",
+        Language.PT: "pt",
+        Language.RO: "ro",
+        Language.RU: "ru",
+        Language.SR: "sr",
+        Language.SK: "sk",
+        Language.SL: "sl",
+        Language.ES: "es",
+        Language.SW: "sw",
+        Language.SV: "sv",
+        Language.TL: "tl",
+        Language.TA: "ta",
+        Language.TH: "th",
+        Language.TR: "tr",
+        Language.UK: "uk",
+        Language.UR: "ur",
+        Language.VI: "vi",
+        Language.CY: "cy",
+    }
+
+    result = BASE_LANGUAGES.get(language)
+
+    # If not found in base languages, try to find the base language from a variant
+    if not result:
+        lang_str = str(language.value)
+        base_code = lang_str.split("-")[0].lower()
+        result = base_code if base_code in BASE_LANGUAGES.values() else None
+
+    return result
+
+
 class BaseWhisperSTTService(SegmentedSTTService):
     """Base class for Whisper-based speech-to-text services.
 
@@ -33,6 +110,7 @@ class BaseWhisperSTTService(SegmentedSTTService):
         model: Name of the Whisper model to use.
         api_key: Service API key. Defaults to None.
         base_url: Service API base URL. Defaults to None.
+        language: Language of the audio input. Defaults to English.
         **kwargs: Additional arguments passed to SegmentedSTTService.
     """
 
@@ -42,11 +120,13 @@ class BaseWhisperSTTService(SegmentedSTTService):
         model: str,
         api_key: Optional[str] = None,
         base_url: Optional[str] = None,
+        language: Optional[Language] = Language.EN,
         **kwargs,
     ):
         super().__init__(**kwargs)
         self.set_model_name(model)
         self._client = self._create_client(api_key, base_url)
+        self._language = self.language_to_service_language(language or Language.EN)
 
     def _create_client(self, api_key: Optional[str], base_url: Optional[str]):
         return AsyncOpenAI(api_key=api_key, base_url=base_url)
@@ -57,6 +137,9 @@ class BaseWhisperSTTService(SegmentedSTTService):
     def can_generate_metrics(self) -> bool:
         return True
 
+    def language_to_service_language(self, language: Language) -> Optional[str]:
+        return language_to_whisper_language(language)
+
     async def run_stt(self, audio: bytes) -> AsyncGenerator[Frame, None]:
         try:
             await self.start_processing_metrics()

src/pipecat/services/groq.py

@@ -11,6 +11,7 @@ from loguru import logger
 
 from pipecat.services.base_whisper import BaseWhisperSTTService, Transcription
 from pipecat.services.openai import OpenAILLMService
+from pipecat.transcriptions.language import Language
 
 
 class GroqLLMService(OpenAILLMService):
@@ -52,8 +53,8 @@ class GroqSTTService(BaseWhisperSTTService):
         model: Whisper model to use. Defaults to "whisper-large-v3-turbo".
         api_key: Groq API key. Defaults to None.
         base_url: API base URL. Defaults to "https://api.groq.com/openai/v1".
+        language: Language of the audio input. Defaults to English.
         **kwargs: Additional arguments passed to BaseWhisperSTTService.
-
     """
 
     def __init__(
@@ -62,11 +63,18 @@ class GroqSTTService(BaseWhisperSTTService):
         model: str = "whisper-large-v3-turbo",
         api_key: Optional[str] = None,
         base_url: str = "https://api.groq.com/openai/v1",
+        language: Optional[Language] = Language.EN,
         **kwargs,
     ):
-        super().__init__(model=model, api_key=api_key, base_url=base_url, **kwargs)
+        super().__init__(
+            model=model, api_key=api_key, base_url=base_url, language=language, **kwargs
+        )
 
     async def _transcribe(self, audio: bytes) -> Transcription:
+        assert self._language is not None  # Assigned in the BaseWhisperSTTService class
         return await self._client.audio.transcriptions.create(
-            file=("audio.wav", audio, "audio/wav"), model=self.model_name, response_format="json"
+            file=("audio.wav", audio, "audio/wav"),
+            model=self.model_name,
+            response_format="json",
+            language=self._language,
         )

src/pipecat/services/openai.py

@@ -54,6 +54,7 @@ from pipecat.services.ai_services import (
     TTSService,
 )
 from pipecat.services.base_whisper import BaseWhisperSTTService, Transcription
+from pipecat.transcriptions.language import Language
 from pipecat.utils.time import time_now_iso8601
 
 try:
@@ -406,6 +407,7 @@ class OpenAISTTService(BaseWhisperSTTService):
         model: Whisper model to use. Defaults to "whisper-1".
         api_key: OpenAI API key. Defaults to None.
         base_url: API base URL. Defaults to None.
+        language: Language of the audio input. Defaults to English.
         **kwargs: Additional arguments passed to BaseWhisperSTTService.
     """
 
@@ -415,13 +417,17 @@ class OpenAISTTService(BaseWhisperSTTService):
         model: str = "whisper-1",
         api_key: Optional[str] = None,
         base_url: Optional[str] = None,
+        language: Optional[Language] = Language.EN,
         **kwargs,
     ):
-        super().__init__(model=model, api_key=api_key, base_url=base_url, **kwargs)
+        super().__init__(
+            model=model, api_key=api_key, base_url=base_url, language=language, **kwargs
+        )
 
     async def _transcribe(self, audio: bytes) -> Transcription:
+        assert self._language is not None  # Assigned in the BaseWhisperSTTService class
         return await self._client.audio.transcriptions.create(
-            file=("audio.wav", audio, "audio/wav"), model=self.model_name
+            file=("audio.wav", audio, "audio/wav"), model=self.model_name, language=self._language
         )
 
 

src/pipecat/transcriptions/language.py

@@ -54,6 +54,9 @@ class Language(StrEnum):
     AZ = "az"
     AZ_AZ = "az-AZ"
 
+    # Belarusian
+    BE = "be"
+
     # Bulgarian
     BG = "bg"
     BG_BG = "bg-BG"
@@ -264,6 +267,9 @@ class Language(StrEnum):
     MN = "mn"
     MN_MN = "mn-MN"
 
+    # Maori
+    MI = "mi"
+
     # Marathi
     MR = "mr"
     MR_IN = "mr-IN"