Add new GPT-4o transcription option to OpenAIRealtimeBetaLLMService

src/pipecat/services/openai_realtime_beta/__init__.py

@@ -1,3 +1,3 @@
 from .azure import AzureRealtimeBetaLLMService
-from .events import InputAudioTranscription, SessionProperties, TurnDetection
+from .events import InputAudioTranscriptionModels, SessionProperties, TurnDetection
 from .openai import OpenAIRealtimeBetaLLMService

src/pipecat/services/openai_realtime_beta/events.py

@@ -16,8 +16,17 @@ from pydantic import BaseModel, Field
 #
 
 
-class InputAudioTranscription(BaseModel):
-    model: Optional[str] = "whisper-1"
+class InputAudioTranscriptionModels:
+    class Whisper1(BaseModel):
+        model: Optional[Literal["whisper-1"]] = "whisper-1"
+
+    class GPT4o(BaseModel):
+        model: Optional[Literal["gpt-4o-transcribe-latest"]] = "gpt-4o-transcribe-latest"
+        language: Optional[str] = None
+        prompt: Optional[str] = None
+
+
+InputAudioTranscription = Union[InputAudioTranscriptionModels.Whisper1, InputAudioTranscriptionModels.GPT4o]
 
 
 class TurnDetection(BaseModel):