Voice Tryouts

A scratchpad for live voice and speech APIs. Each experiment is self-contained and instrumented so the numbers can be compared.

Live transcription over WebRTC

Streams microphone audio to OpenAI's Realtime API with gpt-live-transcribe and benchmarks every stage of the handshake, down to time-to-first-word.

OpenAI Realtimegpt-live-transcribeWebRTCbenchmark

Voice input for AI Elements

Wires real API transcription into an AI Elements PromptInput through a mic button: cross-browser dictation with model choice, delay control and pre-roll, instead of the built-in Web Speech mic.

AI Elementsgpt-live-transcribews-prerolluseVoiceInput