403 Forbidden Error in conversation.item.input_audio_transcription.failed Event on OpenAI Realtime API

API
1

Hello,

I am currently using OpenAI’s Realtime API for audio transcription, and I encountered the following error:

conversation.item.input_audio_transcription.failed
Received event: {
  "type": "conversation.item.input_audio_transcription.failed",
  "event_id": "event_B3IOkncGFKTedHKKQefJ3",
  "item_id": "item_B3IOfpp0N4WQ3ytiWvULi",
  "content_index": 0,
  "error": {
    "type": "server_error",
    "code": null,
    "message": "Input transcription failed for item 'item_B3IOfpp0N4WQ3ytiWvULi'. 403 Forbidden",
    "param": null
  }
}

The conversation functionality itself is working fine. However, I configured the session with input_audio_transcription to capture transcriptions of the input audio. Instead of receiving the expected conversation.item.input_audio_transcription.completed event, I’m getting a conversation.item.input_audio_transcription.failed event with the error.

Could anyone provide insight into why the transcription is failing with a 403 Forbidden error and any steps I might take to resolve this issue?

1 Like
2

Getting the same error forbidden - did you resolve this by chance?

1 Like
3

I have the same problem. I noticed that I get “No eligible models available.” for the user transcript model in the playground as well, but I’m unsure why. Have you found a solution?