Live transcription
Transcription turns the interviewer's audio into text you can read while they are still talking.
How it works
Once a session starts, the transcription panel shows a rolling caption of what is being said. Captions run through cloud speech-to-text for every signed-in account — there is no local engine to pick. It is genuinely useful in two situations: when you miss a word, and when the question is long enough that by the end you have forgotten the start.
Audio sources
| Mode | What it captures |
|---|---|
| System | Audio playing on your machine — the interviewer's voice. The default. |
| Microphone | Your own input. |
| System + Microphone | Both, as separate tracks, so the two sides stay distinguishable. |
The same menu lets you pick the exact devices rather than just the mode: which output the transcriber listens to, and which microphone it uses for your side. By default it follows the OS default devices as they change. Point transcription at the device the interviewer actually comes through when you route the call through a headset or a virtual audio device.
Mock interviews force microphone-only capture — you are the candidate, so the app listens to you. Live copilot can use any of the three modes.
Sending a question to the assistant
- ⌘⌥. — append the latest transcript chunk to your draft without sending it.
- ⌘⌥↵ — send the draft, plus any attached screenshots, to the assistant.
- ⌘⌥, — clear the draft text without removing attached screenshots.
The two-step append-then-send exists so you can stack several sentences — or a transcript line plus a screenshot — into one question.
If detection feels off
Interviewers speak at very different paces. If the app is cutting questions early or lumping two together, adjust the pause interval in settings — a longer interval suits someone who thinks out loud, a shorter one suits rapid back-and-forth.