#no deepgram response

1 messages · Page 1 of 1 (latest)

white kestrel May 20, 2025, 6:46 PM

Hi Deepgram Support,

I’m building a real-time transcription app using Deepgram v3 SDK. My setup:

Frontend (Chrome): MediaRecorder (audio/webm;codecs=opus) sends audio chunks (500ms, ~8kB) via WebSocket to a Node.js backend.

Backend: Streams chunks directly to Deepgram’s listen.live (language: nl).

Issue:
Everything connects and all chunks are sent/received, but I never get any transcriptReceived event or transcript—no errors, no warnings. The Deepgram stream opens, accepts data, and closes cleanly after a linger timeout, but returns nothing.

Tried:

Not sending encoding/sample_rate (per docs for WebM/Opus)

Different chunk sizes

Full event logging (open/data/transcriptReceived/error/close)

Data confirmed received by Deepgram

Questions:

Is streaming Chrome’s MediaRecorder audio/webm;codecs=opus supported with Deepgram v3 for Dutch?

Any known issues or extra requirements for streaming fragmented WebM/Opus?

Would OGG/Opus work better for streaming?

How can I debug why no transcript is returned?

remote shoreBOT May 20, 2025, 6:46 PM

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

Based on the information in the knowledge sources, here’s what I can confirm and recommend regarding your questions:

1. Is streaming Chrome’s MediaRecorder audio/webm;codecs=opus supported with Deepgram v3 for Dutch?

Deepgram does support Opus audio for live streaming transcription, as confirmed by an official answer: “Yes, Deepgram supports Opus 😊” Does live transcription support Opus?. However, the documentation and discussions do not explicitly confirm support for fragmented WebM containers (as produced by Chrome’s MediaRecorder) for live streaming. Most examples and troubleshooting focus on raw Opus frames or OGG/Opus, not WebM/Opus.

2. Any known issues or extra requirements for streaming fragmented WebM/Opus?

There are no explicit statements in the knowledge sources about support or issues with fragmented WebM/Opus for live streaming. However, several discussions highlight that sending audio in an unsupported or mismatched format (container or encoding) can result in no transcription and no errors—just as you’re experiencing. For example, if the encoding or container does not match what Deepgram expects, the connection may open and close cleanly, but no transcript will be returned Web socket connects but does not send back results sporadically, No response received from deepgram streaming websockets, Issue with streaming data not making it to the Deepgram API.

3. Would OGG/Opus work better for streaming?

While there is no explicit recommendation in the sources, it is implied that OGG/Opus or raw Opus frames are more commonly used and tested with Deepgram’s streaming API. If you have the option, try switching to OGG/Opus or sending raw Opus frames, as this may resolve the issue.

4. How can I debug why no transcript is returned?

Check Audio Format: Ensure the audio format and container match what Deepgram expects. If you’re sending WebM/Opus, try switching to OGG/Opus or raw Opus.
Explicitly Set Encoding: Even though the docs say you don’t need to set encoding/sample_rate for WebM/Opus, try explicitly setting them to see if it helps.
Corrupt/Unsupported Audio: If Deepgram receives corrupt or unsupported audio, it will silently fail to return results Web socket connects but does not send back results sporadically.
Test with Known Good Audio: Use a known working Opus or OGG/Opus file and stream it to Deepgram to verify your pipeline.
Use Deepgram’s Example Code: Reference the official Node.js examples for streaming audio to Deepgram, as they may contain subtle implementation details that are important Twilio->Deepgram live streaming in Node.js keeps disconnecting over and over.

Summary

Opus is supported, but fragmented WebM/Opus is not explicitly documented as supported for live streaming.
OGG/Opus or raw Opus frames are more likely to work reliably.
If you get no transcript and no errors, it’s likely a format/container mismatch.
Try switching to OGG/Opus or raw Opus, and use Deepgram’s example code as a reference.

If you continue to have issues, try capturing a short sample of the audio you’re sending and test it with Deepgram’s pre-recorded API to see if it transcribes. This can help isolate whether the issue is with the audio format or the streaming pipeline.

If you need further help, providing a sample audio file and the exact parameters used in your Deepgram request will help the support team debug further.