Compare

OpenAI Realtime

Same talk / ring UX as ElevenLabs. Voice is GPT Realtime; office ambience, keyboard, and mouse are mixed locally and never sent back into the microphone.

AI-HUMAN-VOICE · OPENAI

Ready when you are

Start a browser session with OpenAI Realtime

Status: disconnected

FINETUNE

Voice and model apply on the next call. Mixer, ducking, and VAD update live. Ambience / keyboard / mouse are local Web Audio — never asked of GPT Voice.

Voice Model Speech speed0.96×Office ambience0.12Duck under speech−14 dBKeyboard on tool callsKeyboard intensity0.55Mouse click around toolsVAD silence600 msVAD threshold0.55