Words appear as they're spoken.

Two recordings, streamed through the Saidwire API in real time. Play one and watch the transcript arrive exactly as it did live: first guesses while the person talks, then the final text when they pause.

The first clip is a fictional booking call with synthetic voices. The second is a public-domain recording of President Kennedy at Rice University in 1962, with crowd noise and applause; any word the API got wrong is left as it was.