Free 3 min read
Is Deepgram Down?
Check live Deepgram status — real-time streaming transcription (Nova-3), pre-recorded audio, text-to-speech (Aura), speaker diarization, and API. See recent incidents and set up free email alerts.
Deepgram — live status
Updated every 5 minutes. Full history at prismix.dev/service/deepgram.
Quick check: is Deepgram down right now?
- Prismix: prismix.dev/service/deepgram — live status + 30-day uptime + incidents.
- API call:
curl https://prismix.dev/api/v1/statuses | jq '.services[] | select(.id=="deepgram")' - Direct test:
curl https://api.deepgram.com/v1/listen -H "Authorization: Token YOUR_KEY" --data-binary @audio.wav— a 200 response confirms the API is reachable.
Common causes of "Deepgram not working"
- WebSocket connection closes immediately — Deepgram's real-time API uses WSS; ensure you're sending the Authorization header (not a query parameter) and that your audio format parameters (encoding and sample_rate) in the URL exactly match the audio you're sending. A format mismatch causes a silent immediate disconnect with no error message.
- "Transcript is empty" or no results returned — for streaming, verify you're sending audio in the correct encoding (linear16, mulaw, etc.) at the sample_rate you specified. Deepgram returns partial transcripts via streaming events, not a single bulk response — you must implement an onMessage handler that listens for the Results event type to receive transcript data.
- Nova-3 model not available on your account — Nova-3 is only available on select paid plans; check your current plan at console.deepgram.com. Use model=nova-2 as a fallback, which is available across all plan tiers and still provides high accuracy for most use cases.
- Text-to-speech (Aura) voice not recognized — Aura voice names are case-sensitive: use the full format such as aura-asteria-en or aura-luna-en. Voice names changed between the Aura beta and the current release; if you receive a 400 error, check the current voice name list in the Deepgram docs as your voice name format may be outdated.
- Diarization producing too many speakers — Deepgram diarization auto-detects speaker count and does not accept an exact expected speaker count parameter (unlike some other services). For cleaner diarization with a known speaker count, you can experiment with the diarize_version parameter, but results depend on audio quality and speaker separation.
- Credit balance exhausted (HTTP 402) — Deepgram charges per second of processed audio; pre-recorded and streaming transcription have different per-second rates. The API returns HTTP 402 when your balance reaches zero. Check your current balance and usage breakdown at console.deepgram.com before troubleshooting other errors.
Set up free email alerts for Deepgram
- 1
Sign in to Prismix
Go to prismix.dev/sign-in — email OTP or GitHub sign-in.
- 2
Star Deepgram
On prismix.dev/service/deepgram, click the ☆ star icon.
- 3
Alerts are live
You'll get an email within minutes of any status change.
🔔
Stop manually checking — get alerts instead
Star Deepgram on Prismix and get emailed the moment status changes. Free, no credit card.
Monitor other AI audio & speech tools?
Full status dashboard: prismix.dev/status