#WebSocket real-time transcription doesn't send transcription back even tho connection opened

1 messages · Page 1 of 1 (latest)

nocturne ether
#

Hi Deepgram Support Team,

We’re seeing an issue where the WebSocket real-time transcription connects successfully, but no transcription text comes back. The audio stream is sent correctly, but the responses stop returning intermittently. Restarting sometimes fixes it, but it’s inconsistent.

Here’s a short Loom video showing the issue: Loom video link

Could you let us know if this is a known problem, and the best way to debug or handle cases where the stream stays open but no results are sent?

Thanks for your help!

Best,
Alex

languid cradleBOT
#

Hey there! This channel allows you to ask Kapa.ai for assistance. But it looks like you haven't connected your Discord profile to your Deepgram account yet. You can do so at https://community.deepgram.com - by doing so, you'll be able to ask questions in #1115960287183511643 if you need human assistance!

hollow sparrowBOT
#

Hi @nocturne ether. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

Hi Alex,

This is a known issue that can occur for several reasons when using Deepgram’s real-time transcription WebSocket API. Based on your description and the knowledge sources, here are the most common causes and recommended debugging steps:

Common Causes

  1. No Audio Data or Corrupt Audio Sent

    • If the WebSocket connects but Deepgram receives no audio data, or the audio is in an unsupported or corrupt format, the connection may remain open but no transcription will be returned. This is confirmed by Deepgram staff in several reports, where request logs showed either no audio or corrupt/unsupported audio as the root cause. Double-check your audio encoding, sample rate, and that you are actually sending audio data after the connection opens. For example, in one case, Deepgram support found that "we received corrupt or unsupported audio data in that stream, causing us to close the stream" and no results were returned Web socket connects but does not send back results sporadically.
  2. KeepAlive and Data Flow

#
  1. Audio Format Mismatch

    • Ensure that the encoding parameter in your WebSocket connection matches the actual audio format you are sending. Mismatches can cause Deepgram to silently drop the audio or close the stream without results Using Lower-Level Websockets with the Streaming API.
  2. Network or Service Instability

    • There have been periods of high load or incidents on Deepgram’s side that caused intermittent connection or latency issues. If you notice the connection is stuck in "connecting" or results are delayed, retrying the connection after a short wait (1–2 seconds) is recommended. Deepgram suggests waiting up to 5 seconds for a connection and retrying a few times, as different datacenters may be affected differently Live Transcription Ready State Connection Latency.

Debugging Steps

#
  • Check Request IDs and Deepgram Console: For each session, log the Deepgram request ID and compare your logs with the Deepgram Console. If you see "0 audio duration" or similar, it means no audio was received Streaming STT with no transcription and zero audio duration.
  • Validate Audio Stream: Ensure your audio stream is not muted, empty, or in an unsupported format. Avoid sending empty bytes (b''), as this can cause unexpected closures STT Troubleshooting WebSocket, NET, and DATA Errors.
  • Handle Reconnection Gracefully: If the connection drops or hangs, close the socket and reinitialize a new connection. Buffer audio data during reconnection to avoid data loss Recovering From Connection Errors & Timeouts When Live Streaming.
  • Check for Errors on Deepgram Side: If possible, share the request IDs with Deepgram support for deeper investigation, as they can check for backend errors or audio issues.

Summary

This is a known, intermittent issue often related to audio data not being received, audio format problems, or network/service instability. Double-check your audio stream, encoding, and keepalive logic. If the problem persists, collect request IDs and logs and share them with Deepgram support for further analysis.

Let me know if you need more specific troubleshooting steps or want to share a request ID for a deeper look!

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@nocturne ether Kapa.ai is still learning and improving, please let me know how I did by reacting below.