Realtime Voice Lab
Session not started
Checking model...
User
Speak in English, Chinese, Malay, Thai, Indonesian, Japanese, or mixed language.
Assistant
The spoken answer and transcript will appear here.
Latency
Turn timing
| Turns | First Token Latency | Response Complete Latency |
|---|
Metric Definition
First Token Latency is time to first model output. Response Complete Latency is time until the model finishes the full spoken reply.
Configuration Prompt, voice, transcription, VAD
Behavior
Prompt studio
Prompt notes
Session
Audio setup
Transcript model: gpt-realtime-whisper
Saved session
Conversation log
Transcript model: gpt-realtime-whisper