You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm establishing a WebSocket connection to the ElevenLabs Conversational AI endpoint, sending Twilio audio in real time (user_audio_chunk in base64). The agent_response come through almost immediately, but within the first 5-10 seconds the user_transcript is consistently missed. After that, transcripts arrive as expected.
It was working great last week, and I haven't made any changes to my code.
I then converted audio format to PCM 8000 Hz to "Test AI agent" on the ElevenLabs dashboard, and it is having the same issue. First ~5 seconds is ignored.
Finally, I went to the API Reference and started a chat with ElevenLabs "Need help?" assistant widget, and had the same issue there.
Is anyone else experiencing this delay?
Logs:
2025-01-22 14:54:44.846 [INFO] [Twilio] Twilio connected to media stream.
2025-01-22 14:54:45.698 [INFO] [11Labs] Connection fully established and ready for audio
2025-01-22 14:54:45.872 [INFO] [11Labs] Received agent response
// MISSING FIRST USER TRANSCRIPT HERE
// BUT GET THE NEXT ONE 12 SECONDS LATER
2025-01-22 14:54:58.053 [INFO] [11Labs] Received user transcript
2025-01-22 14:54:58.212 [INFO] [11Labs] Received agent response
2025-01-22 14:55:02.855 [INFO] [11Labs] Disconnected from Conversational AI.
2025-01-22 14:55:03.828 [INFO] [Twilio] Client disconnected.
The text was updated successfully, but these errors were encountered:
patricktargun
changed the title
First user transcript delayed by ~5 seconds using real-time WebSocket API
user_transcript missed in opening ~5 seconds with WebSocket API
Jan 23, 2025
patricktargun
changed the title
user_transcript missed in opening ~5 seconds with WebSocket API
user_transcript missed in opening ~5 seconds
Jan 23, 2025
I'm establishing a WebSocket connection to the ElevenLabs Conversational AI endpoint, sending Twilio audio in real time (user_audio_chunk in base64). The agent_response come through almost immediately, but within the first 5-10 seconds the user_transcript is consistently missed. After that, transcripts arrive as expected.
It was working great last week, and I haven't made any changes to my code.
I then converted audio format to PCM 8000 Hz to "Test AI agent" on the ElevenLabs dashboard, and it is having the same issue. First ~5 seconds is ignored.
Finally, I went to the API Reference and started a chat with ElevenLabs "Need help?" assistant widget, and had the same issue there.
Is anyone else experiencing this delay?
Logs:
The text was updated successfully, but these errors were encountered: