Yes. An AI interview assistant can transcribe both the interviewer and you because the two voices arrive through two different inputs: the interviewer comes out of your computer's audio output, and you go in through the microphone. A desktop assistant reads both streams and labels each line by where it came from, so the transcript shows what was asked and what you answered.
The separation is clean with headphones and gets muddy with speakers, because the interviewer's voice then leaks into your microphone. Here is how the two streams work, what breaks them apart, and what a two-sided transcript is good for afterward.
How one assistant transcribes both the interviewer and the candidate
Everything the meeting app plays, including the interviewer, comes out of one channel: the computer's audio output. A desktop assistant captures that system audio without joining the meeting. Your own voice never passes through that channel; it comes in through the microphone, which the assistant reads as a second stream.
Speaker separation in a desktop assistant is therefore done by input, not by voice: system audio is the interviewer, the microphone is you. That is reliable in a one-on-one call. In a panel interview every interviewer shares the system-audio stream, so the transcript knows "interviewer side" and "you" but not which panelist spoke; we explain the limits in speaker separation in interview AI.
Why headphones keep the streams apart and speakers do not
With headphones, the interviewer's voice goes only to your ears, and the microphone hears only you. With speakers, the interviewer's voice fills the room and the microphone hears it too. The same sentence then arrives twice: once cleanly from system audio and once as a muffled echo on your microphone track. Captions duplicate, your side of the transcript fills with the interviewer's words, and question detection can fire on the echo. We tested both setups in headphones vs speakers; wired headphones or earbuds gave the cleanest split.
| Setup | Interviewer transcript | Your transcript | Risk |
|---|---|---|---|
| Wired headphones or earbuds | Clean | Clean | Low |
| Bluetooth earbuds | Clean, slightly delayed | Clean | Low; captions lag a little |
| Laptop speakers and built-in mic | Clean | Mixed with an echo of the interviewer | Duplicate captions, false question detection |
| External speakers in a room | Clean | Heavily mixed | Your side of the transcript is unusable |
What breaks the separation
- Speakers. The bleed described above. Switch to headphones.
- The wrong output device. If the meeting app plays through a monitor's speakers while the assistant listens to the default device, the interviewer goes missing. Set the meeting app and the assistant to the same output.
- Audio enhancement software. Some vendor tools route the microphone back through the system output, which makes you appear on the interviewer's track. Turn those off for the call.
- A second person in the room. Anyone near the microphone lands on your track.
- Microphone permission. On macOS the assistant needs microphone permission; without it your side is silent.
A thirty-second check before the call catches all of these: say one sentence while a video plays, and confirm the two appear on different sides.
What a two-sided transcript is good for afterward
During the call, the interviewer's side drives the captions and the suggested answer. After the call, your side is the valuable half. You can reread exactly what you said, count filler words, compare the length of each answer to the question that was asked, and spot where you answered a different question from the one posed. Reviewing your own interview walks through a fifteen-minute routine.
With CoPilot Interview, audio is streamed in short segments for transcription and is not stored on our servers; a text transcript is saved only in your Documents folder, and the transcribed text goes to the AI model you choose. The transcript is a plain file you can read, copy or delete.
Where CoPilot Interview fits
CoPilot Interview is a desktop AI interview assistant that listens to the computer's audio output for the interviewer and to your microphone for you, so it works the same way with Zoom, Microsoft Teams, Google Meet, Webex and Amazon Chime and never joins the meeting. It shows live captions of each question and a structured answer in about 2 seconds. Before any live use, the AI mock interviews read questions aloud and score each spoken answer out of 10, which is also the simplest way to see a two-sided transcript. Live help is for interviews where it is allowed, so follow your employer's rules.
Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.
Frequently asked questions
Does the assistant need to join the meeting to hear the interviewer?
No. The interviewer's voice is already coming out of your computer's audio output, and a desktop assistant reads that stream directly. It never connects to the meeting, so it is not a participant and needs no invitation or admit prompt.
Can it tell two interviewers apart in a panel?
Not by voice. Every interviewer arrives on the same system-audio stream, so the transcript labels that side as the interviewer and your microphone as you. Who spoke within the panel is not identified. You can add names yourself when you review the transcript afterward.
Is my voice recorded and kept somewhere?
No audio file is kept. Audio is streamed in short segments for transcription and is not stored on our servers. What is saved is a text transcript in your Documents folder, which includes your words as text, and nothing is uploaded from that folder.
Do Bluetooth earbuds work?
Yes. Bluetooth adds a little delay, so captions can lag the spoken question slightly, but the separation between the interviewer and you stays clean because the interviewer's voice never reaches your microphone.
Why do I see the interviewer's words on my side of the transcript?
Almost always speakers. The interviewer's voice is leaving your speakers and entering your microphone, so it is transcribed twice. Switch to headphones or earbuds and the duplicate lines stop. Audio enhancement software that routes the microphone through the system output can cause the same effect.
Hear both sides in a mock interview
The AI interviewer reads questions aloud and scores your spoken answers, free.
Start a mock interview