I’m experiencing an issue with the OpenAI Realtime API transcripts when using a custom websocket server.
So far I have:
-
Custom Node.js relay server (instead of direct browser-to-OpenAI connection)
-
Using @openai/realtime-api-beta package
-
Audio streaming and speech synthesis working correctly
When using the relay server, I’m not receiving the text transcripts in the response, although audio streaming works fine. However, when I connect directly to the OpenAI API from the browser (without the relay), I receive both audio and text transcripts correctly.
Anyone else experience an issue like this?