Following the official docs from https://platform.openai.com/docs/guides/speech-to-text it says “Below is an example payload for setting up a transcription session:”. I am using "type": "transcription_session.update" as it says. However when sending that config through the WebSocket to the URL "wss://api.openai.com/v1/realtime?intent=transcription" it gives me an error and won’t transcribe.
{‘type’: ‘error’, ‘event_id’: ..., ‘error’: {‘type’: ‘invalid_request_error’, ‘code’: ‘invalid_value’, ‘message’: “Invalid value: ‘tra…ate’. Supported values are: ‘session.update’, ‘input_audio_buffer.append’, ‘input_audio_buffer.commit’, ‘input_audio_buffer.clear’, ‘conversation.item.create’, ‘conversation.item.truncate’, ‘conversation.item.delete’, ‘conversation.item.retrieve’, ‘response.create’, and ‘response.cancel’.”, ‘param’: None, ‘event_id’: None}}
The error fails to even mention transcription_session.update as an option. I am trying to use the gpt-4o-transcribe model btw in my Python script. Right after connecting to the websocket (with websockets package) I run this: await stt_ws.send(json.dumps(STT_CONFIG)), with STT_CONFIG being very similar to the one from the documentation.
Anybody know what is going on and how to either get this code from the docs working or perhaps some different payload to get transcription working?