Free 3 min read

Is Hume AI Down?

Check live Hume AI status — EVI WebSocket, speech prosody, and empathic audio models. See recent incidents and set up free email alerts.

Hume AI live status

Hume AI — live status

Updated every 5 minutes. Full incident history at prismix.dev/service/hume.

Full status →

Quick check: is Hume AI down right now?

  1. Prismix: prismix.dev/service/hume — live status + 30-day uptime + incidents.
  2. Hume AI status: prismix.dev/service/hume — Hume AI has no dedicated status page; Prismix is the best monitor.
  3. API call: curl https://prismix.dev/api/v1/statuses | jq '.services[] | select(.id=="hume")'

Set up free email alerts for Hume AI

  1. 1

    Sign in

    Go to prismix.dev/sign-in — email OTP or GitHub sign-in.

  2. 2

    Star Hume AI

    On prismix.dev/service/hume, click the ☆ star icon.

  3. 3

    Alerts are live

    You'll get an email within minutes of any status change.

Common causes of "Hume AI not working"

If Prismix shows Hume AI as "Operational" but you're having issues:

  • EVI WebSocket disconnecting unexpectedly — EVI uses a persistent WebSocket session. Network interruptions, browser tab backgrounding, or missing keepalive pings cause silent disconnects. Implement a reconnection loop with exponential backoff and resend the session_settings message after each reconnect.
  • Prosody model latency spike (>2s response) — the speech prosody pipeline involves audio encoding, model inference, and TTS synthesis. Latency spikes usually occur during high demand. Reduce audio chunk size to under 100ms to keep the streaming pipeline flowing and monitor RTF (real-time factor) in the response metadata.
  • API key returning 401 Unauthorized — Hume AI API keys are workspace-scoped. Ensure the key is from the correct workspace at platform.hume.ai and is passed in the X-Hume-Api-Key header, not as a Bearer token.
  • Audio input format rejected — EVI requires 16kHz mono PCM audio. Common rejections: 44.1kHz stereo from browser MediaRecorder, MP3 or OGG encoding, or float32 instead of int16 PCM. Use the Web Audio API to downsample and convert to the correct format before sending.
  • Turn-end detection false positive interrupting speech — EVI's voice activity detection (VAD) may cut the user short in noisy environments or with slow speakers. Adjust the vad_sensitivity parameter in your session_settings to a lower value (e.g. 0.3) to require longer silence before declaring end-of-turn.
  • Enterprise SSO authentication failure — enterprise accounts using SAML SSO must complete IdP setup in the Hume dashboard before API keys are valid. If SSO was recently enabled or an IdP certificate was rotated, API keys may be invalidated and need to be regenerated from the platform.hume.ai settings page.
🔔

Stop manually checking — get alerts instead

Star Hume AI on Prismix and get emailed the moment status changes. Free, no credit card.

Monitor related voice and audio AI tools?

Full status dashboard: prismix.dev/status