AI for interviews: a footing, not a script
A question lands once, and you have seconds. Whisperer follows the whole conversation and puts up a footing — the shape of an answer and your own material from your knowledge base. The words stay yours.
60 free minutes a month · macOS and Windows
The answers below are an example. Your own base connects in the app.
…
…
Let's start with Q3 — how did conversion move?
Conversion is up 14% — let me pull up the funnel.
And the cost per lead?
Cost per lead is €12.40 — 9% under plan. Most of the gain came from organic.
Not a video — this is Whisperer itself, and the panel really answers. The microphone starts only when you press it.
The question
What happens while you take a breath
Help in an interview is worth something in exactly one window: after the question lands and before you start talking. Anything that arrives later arrives too late.
- Them The question is asked Whisperer listens to two streams: the other side through system audio, you through the microphone. Turns are labelled by speaker.
- Whisperer It reaches the transcript The line runs in the window above your screen. That same text becomes the question if you press enter on an empty field.
- Your notes Your context is pulled in Your own notes are mixed into the answer: projects, numbers, stories. Without them the prompt is generic, and generic sounds borrowed.
- You A footing appears Not a line to deliver — an order and your own material. There is nothing to read out, and reading out is audible anyway.
Limits
Two things worth knowing now, not a minute before the call
Both follow from how the product is built, and neither is a setting you can flip. If either one rules out your case, better to find out before you sign up.
A finished recording can be uploaded: the Meetings section takes one file up to 25 MB — mp3, m4a, wav, ogg, flac, aac, plus mp4 and webm, from which the audio track is used. What is missing is bulk upload of a whole archive; and it spends the same minutes a live call does.
The other side is heard through system audio: on macOS that means granting the Screen Recording permission, on Windows no extra permission is needed. In a physical meeting room with no online call, where only a microphone is available, the transcript will contain you and no one else.
The line
Help and substitution are not the same thing
The line does not run along the technology. It runs along one question: does the tool replace the skill being examined right now? We build an assistant for conversations, so we will say it plainly — there are things Whisperer will not do, and things that are not worth doing.
- Your own notes in front of youProject numbers, stack, names. Recall for figures is not the skill anyone is testing.
- The shape of an answerA reminder of the order: situation, action, result. The substance is still yours.
- Wording in a second languageEspecially once you have said so out loud — one sentence turns a grey area white.
- A model answering about your experienceThe question was about you. The first follow-up exposes it, and sometimes it surfaces later and costs more.
- Solving the task instead of youLive coding examines how you think out loud, not the final code.
- Staying quiet when the rules were statedIf they asked, lying is out. An agreement outranks convenience.
The long version, with cases, is in AI in interviews: where help ends and cheating begins. The position there is the same as here.
Before
Three things before the call
A prompt is only as good as your context. An empty knowledge base produces generalities, and that is the most common reason people are disappointed.
Fill the knowledge base
Five to ten notes: your role, your stack, projects with outcomes, two or three stories for the usual competencies. The knowledge base opens on paid plans.
Start the session before the call
The window sits above your desktop. There is a setting that keeps it out of screen capture — turn it on in advance, not when someone asks you to share.
Make sure both sides are heard
Both speakers should appear in the transcript. If you only see your own lines, system audio access is missing and the prompt is answering into a void.
Questions
What people ask first
Can the interviewer see the window when I share my screen?
A window excluded from capture by the operating system does not appear in the shared stream, in recordings or in screenshots. That protects against software capture only: a camera, a person next to you, a phone pointed at the screen, and your own behaviour are not covered. On Windows this requires version 10 build 2004 or newer.
Does Whisperer answer for me?
No. It shows a footing: the order of an answer and your own facts from the knowledge base. Reading it out is pointless — it is audible, and the first follow-up question exposes it. The product does not promise you will pass an interview, and could not.
Can I upload a finished recording?
Yes. In the Meetings section you pick the file with a button or drag it in — one at a time, up to 25 MB: mp3, m4a, wav, ogg, flac, aac, plus mp4 and webm, from which the audio is taken. From there it goes exactly like a live call: transcript, meeting map, decisions and tasks. A whole archive cannot be uploaded in one go, and in no-logs mode upload is unavailable — nothing is stored there by design.
Will proctoring software notice an assistant is running?
It may. Proctoring is usually installed on your machine and looks at the process list rather than at the picture on screen, so an uncapturable window does not help. Where the rules of an assessment have been stated, they apply.
Do I need to connect Zoom or Google Meet?
No. Whisperer is a window above your screen, not a bot in the call and not a plugin for the conferencing service. It does not care where the conversation happens: Zoom, Google Meet, Microsoft Teams, Telegram, Discord or anything else.
What does it cost to try?
The Free plan gives 60 minutes of recognition a month. Start, Pro and Max lift the minute limit; the knowledge base and the advanced prompt modes open on the paid plans.
Is the conversation stored?
By default the transcript is kept in your account history. If you would rather keep nothing, switch on no-logs mode: the transcript and the answers are not saved after the session ends, and minutes are consumed as usual.
Try it on one conversation
Not an important one. Ask a colleague for fifteen minutes, put three questions to each other, and watch whether the footing matches what you were actually asked. That is enough to know whether this is your tool.