Using the app

First run

1. Open the app and sign in (see Sign in to the desktop app).
2. With an active subscription, the app opens on the Session page. Without one it shows a card, Upgrade to use AI Interview Assistant, with an Upgrade button that opens the upgrade page on this website. Its pages stay hidden until your plan is active (see Subscribe).
3. Set up a session (below), then press Start session. There is no setup wizard.

The left sidebar has the pages Session, Routing, Providers, Prompt profiles, Settings and Account. At the bottom of the sidebar, a dot and a word show what the session is doing (Idle, Starting, Listening, Paused, Generating, Degraded, Stopping, Stopped or Faulted).

Set up a session

The Session page has a setup panel on the left (the toggle icon at the top, with the tooltip Show or hide the setup panel, collapses it) and the live view on the right. Nothing is recorded or sent while you set up; Preview, below, checks a single frame on your computer.

What to listen to

• Call audio (desktop output). Choose the output device your call plays through. It lists your playback devices with the default one preselected. The app captures everything that device plays: the call, and also music, videos and notification sounds. Mute or close other audio during a session.
• Microphone. A checkbox, off by default, with its own device list. In the transcript your voice is labeled "You (microphone)" and the call "Interviewer (call audio)".
• Speech language. A language code such as en-US. The default is en; empty uses the recognizer's default.

Refresh rereads the device list. It also checks for changes every few seconds. If the app finds no output device it says "No desktop audio devices were found. Check that an output device is enabled."

What to read on screen

In the Screen card, Capture is Off by default. The other choices are Window, Monitor and Region.

• Source lists the windows (with their size) or monitors you can pick. Refresh rereads the list.
• Region (x, y, width, height) optionally limits reading to part of the source, in pixels. Clear removes it.
• Preview captures one frame to check what the app can read. It shows only a status line such as "Preview: 1920×1080 at 40 ms.", not a picture.
• OCR language is the language of the text on screen. The default is en-US.

The app reads the chosen window with text recognition on your computer. The text it finds goes with a question when it is relevant. Pictures of your screen are not sent to the service.

You need at least one source before you start: the call audio device, the microphone or a screen source. Otherwise the panel says "Choose the call audio device, the microphone or a screen source before starting."

How answers are written

• Mode (Practice, AuthorizedInterview, Meeting or Interviewer) names what the session is for. It is recorded with the session and does not change how guidance is produced.
• Prompt profile sets how answers are written. The default is Technical concise. See Prompt profiles.
• Answer length: UltraConcise, Concise (the default), Normal or Detailed.
• Answer language: leave it empty to answer in the language of the question.

Spoken answers

In Spoken answers, Output chooses TextOnly (the default), AudioOnly or TextAndAudio. For audio, pick an Output device and press Test to hear a short sound. Volume and Mute are there too, and Interrupt stale audio (on by default) stops a spoken answer that is out of date when a new question arrives.

Before you start

The Before you start card checks your choices as you make them (Check runs the checks again). It lists anything that blocks starting, and a section called WHERE YOUR DATA GOES that spells out what is sent where. Where the app needs your agreement, it shows a checkbox to tick, for example that questions, recent transcript lines and screen text go to AI Interview Assistant servers, which use an approved external AI provider to generate the answer, and that the content is not stored. Start is available once the checks pass and every box is ticked. See What leaves your computer.

Start, pause and stop

• Start session begins capture. Nothing is recorded before you press it. Check reruns the pre-start checks.
• Pause stops every capture device until you press Resume.
• Stop ends the session and releases every capture device.
• Overlay opens the always-on-top overlay (see below).

The status next to the buttons says what is happening, for example "Ready to start.", "Listening. Questions are detected automatically from the call audio and the shared screen.", "Paused — capture stopped" and "Stopped". While the session runs, the capture summary in Before you start shows whether screen, call audio, microphone, AI and speech are on. You can edit your context while paused, and press Apply context changes to the paused session.

How guidance appears

• Questions are detected automatically, from the call audio and from the shared screen. There is no button to push before asking.
• Detected question shows the question, with its kind (technical, coding, system design, behavioral, follow-up or general), a confidence figure and where it came from (call audio, screen, both, or typed). Before any question it says "No question yet."
• Guidance streams in as the AI writes it. It says "Guidance streams here once a question is detected." until then. The answer is brief and shaped by your prompt profile and answer length.
• Transcript shows the latest lines recognized from the call audio and the microphone, ending with an ellipsis while a line is still being recognized.
• Screen context shows the text read from the shared screen.

Controls on a detected question and its answer:

• Regenerate asks for a new answer to the same question.
• Not a question drops the detected text and its answer. Repeats of it are not answered again.
• Pin keeps a piece of text in every request for the rest of the session (with an empty box it pins the current question). Clear pinned removes the pins, or use the X on a single pin.
• Stop answer stops the answer that is being written. Stop audio stops the spoken answer that is playing.

If detection is not working, open Fallback: type a question and use Ask (fallback). It is meant for accessibility, or for when capture is degraded.

The overlay

Overlay on the Session page opens a small window that stays on top of other windows and is not in the taskbar. It shows the current question ("Waiting for a question…" until there is one) and the streamed guidance. It opens only when you press the button, and it needs an active subscription.

• Move it by dragging its header, or press Alt with the arrow keys; add Shift for larger steps.
• Its position and size are remembered, and it is kept on a connected monitor.
• The header buttons pause or resume the session, stop it, stop audio, open the main window, reset the position and size (back to the top-right corner of the main monitor), and close the overlay. Closing the overlay does not stop the session.

These Alt-arrow keys work in the overlay only. The app defines no global keyboard shortcuts.

Speech

Speech has two jobs: turning the audio you chose into text (transcription), and reading guidance aloud.

• Through the service (what the app uses on every current plan). Short segments of audio from the sources you chose are sent to AI Interview Assistant, which relays them to an approved external speech provider and returns the text. To read guidance aloud, its text is relayed the same way and the audio comes back. Neither the audio nor the text is stored. The session shows Speech · Included. The app asks you to tick a box for it before you start, because call audio can include other people's voices.
• On your computer. Windows' own speech recognition and voices, with nothing uploaded. In newer versions of the app, when speech through the service is refused, for example because the monthly speech allowance is used up, the app tells you and uses them instead (for the rest of the session when the refusal is lasting, or until you pause and resume when it is temporary). They need a speech recognizer or voice installed in Windows for your language; if one is missing, the app says speech cannot be transcribed or read aloud.

There is no switch in the app to choose between them. Your plan includes a monthly speech allowance of up to 30 hours of transcription and 200,000 characters read aloud, separate from AI credits. The Account page shows how much you have used and when it resets. Call audio can include other people's voices: use speech features only where they have been told and, where required, have agreed.

AI included in your plan

• AI guidance comes with your plan. You do not choose an AI provider or enter an API key: Providers shows only a card saying AI is included, and Routing has nothing to set for the language model (its other rows, for speech recognition, vision and spoken output, are not used while speech is included in your plan).
• The Model lists on the Session and Account pages are informational: the service picks the model whichever entry you choose.
• The session shows AI · Included and, after the first answer, how many credits are left. The Account page says "AI credits: … left this month … Resets …" and warns when less than 10% is left.
• During a session, a message tells you when guidance stops: for example "This month's AI allowance is used up. It resets when your next monthly period starts. The Account page shows the date." or "AI guidance is temporarily unavailable. Try again in a few minutes." The Account page shows the same news in its own words, such as "You've used this month's AI allowance. It resets on …".

More about credits, and where to see your usage, is in AI credits and usage.

Prompt profiles

A prompt profile sets how answers are written: the system instructions, the response format and the default length. These are built in:

On the Prompt profiles page you can Duplicate a profile, edit your copy (name, purpose, system instructions, response format and verbosity), Save it and Preview it with a sample question. Your own profiles can be deleted; built-in profiles never change and cannot be deleted. Saved profiles appear in the session's Prompt profile list with "(custom)" after the name.

Your context

In the setup panel, Your context (optional) lets you give the app facts it should use. Everything is optional. Facts are used exactly as you write them and are never invented or expanded, and pasted text is treated as information, never as instructions to the app.

• Each field has a Use checkbox: when it is off the text stays in the editor but is not sent. At most 12,000 characters of context are sent with a question.
• Preview what is sent shows exactly the text that goes with each question.
• After this session chooses what happens to it: Do not save (the default, kept in the editor until you close the app), Use for this session only, or Save as a reusable profile, which stores it on this computer when you press Save profile. Saved profiles can be opened, updated, duplicated and deleted.

Saved history and local data

• Under Settings → Privacy and persistence, Persist guidance (enables History) and Persist transcript (opt-in) are off by default. Turn them on and the app saves detected questions and final guidance, and optionally final transcript lines, in a database on your computer. Raw audio and screenshots are never saved.
• Retention (days) sets how long saved sessions are kept. The default is 30, and the range is 1 to 365.
• The current version of the app has no screen for browsing or exporting saved history.
• Settings → App data locations shows where the app keeps its data, logs and database. Deleting your cloud account does not delete them, and the app has no button that removes them (Clear logsdeletes only the log files).

Settings

Some versions of the app have a Save settings button; newer ones apply and save a change as soon as you make it. Reset to defaults restores the defaults.

• Appearance: Theme is Dark (the default), Light, or Match Windows. High-contrast mode follows Windows.
• Window and notification area: Minimize AI Interview Assistant to the system tray is off by default. While it is off, closing the window during a session asks before stopping (or stops and exits without asking, if you choose that). While it is on, closing during a session always offers to minimize, stop and exit, or cancel, and a second setting chooses what closing does when no session runs: Always ask (the default), Minimize to tray, or Exit. See Closing, the tray icon and signing out.
• Privacy and persistence: the history switches and retention above.
• Advanced: Use system proxy for remote providers, Log detail (Standard is recommended), Open log folder and Clear logs. The log records app events and error categories with credentials redacted, starts a new file each day (or at 10 MB) and keeps at most 7 files.
• About: the app version.

Closing, the tray icon and signing out

• Closing during a session. By default the app asks before stopping: it asks "Stop session and exit?" and you choose Stop and exit or Keep session running. Stopping ends all capture.
• The notification-area (tray) icon is there whenever the app runs. Click it to open the window, or right-click for the menu: open the app, pause, resume or stop the session, stop audio, open Settings, or exit. With Minimize AI Interview Assistant to the system tray on, closing the window offers to minimize to the tray, and a session keeps running there. Minimizing or restoring never starts or stops capture.
• One window. Opening the app again brings the running window to the front.
• Signing out. On the Account page, Sign out stops any session, clears your account sign-in on this computer and keeps your settings and local data. If a session is running it asks first: "Sign out and stop the session?"

The Account page also has Manage subscription and View invoices (both open the billing portal in your browser), Refresh, Upgrade, and Your data with Export account data and Delete account (see Privacy and your data).

Updates

The Microsoft Store delivers updates. If the app shows Update required ("This version of AI Interview Assistant is no longer supported. Update the app to continue."), update it from the Store.