Practice & Assistance/AI Interview Assistant (Web)

AI Interview Assistant (Web)

Use the browser-based assistant to start a session, capture live transcripts, generate answers, stop billing, and save records.

What this page is for

The web assistant is designed for quick starts and lightweight validation. The core flow is: start a session, capture live transcripts, generate answers, stop the session, and review the saved record.

The current web experience is focused on interview voice practice. For coding tests, high-frequency practice, and full control over the environment, use the desktop client instead.

Before you start

Open the download page first, then enter the web experience from there.

Before starting a session, confirm that:

  • You are signed in.
  • The browser has microphone permission.
  • System audio sharing is available if you need to capture the interviewer's voice.

The "system audio" toggle only starts the browser sharing flow. Whether audio is actually captured depends on the audio checkbox in the browser sharing dialog.

Standard workflow

  1. Configure the input source and answer settings, then start the session.
  2. During the session, watch the transcript area and answer area update in parallel.
  3. Use auto answer, general answer, coding answer, or "answer this item" depending on the pace of the conversation.
  4. Stop the session when finished, then wait for billing and record status to refresh.

Answers and settings

The common settings match the AI assistant guide. Start with answer language, speech model, auto answer, and knowledge base.

Auto answer works well for continuous Q&A. When you need tighter control, use general answer, coding answer, or answer a specific transcript item manually.

General answer vs coding answer

ActionBest forRecommendation
General answerSelf-introduction, project experience, behavioral questions, open-ended follow-upsUse it when you need a structured spoken answer quickly. Focus on logic, priorities, and delivery order.
Coding answerAlgorithm questions, implementation tasks, complexity analysis, edge-case follow-upsUse it when you need an executable approach or code outline that can be explained step by step.

Web shortcuts

  • Shift + M: Start or stop the session
  • Shift + G: Generate a general answer
  • Shift + J: Generate a coding answer

Shortcuts do not trigger when the cursor is inside an input or textarea. Move focus away from the input first. You can also open the shortcut help inside the page.

Auto answer trigger conditions

Auto answer runs only when all of these are true:

  • Auto answer is enabled.
  • The session is active.
  • New interviewer speech has been recognized and saved as transcript text.
  • No other answer is currently being generated.

If you manually click general answer, answer this item, or coding answer, the manual action takes priority.

For the first validation run, turn off the knowledge base and test only the basic audio pipeline. Add knowledge base and extra requirements after the voice flow is stable.

Billing, records, and session recovery

After you stop the session, voice usage is settled. Record upload and billing are checked independently.

A record is uploaded only when the session meets the minimum duration and transcript requirements. If those conditions are not met, the session can still stop successfully without creating a full record.

If the same account already has an active session, the page will ask whether to recover it or create a new one. A stopped session cannot accept new answers.

"Session stopped" does not guarantee that a record was uploaded. Use the records page as the source of truth.

Current limits

Stealth mode, keyboard interception, and the full coding-test workflow are desktop-client capabilities. For long sessions or frequent practice, switch to the desktop client.

If your goal is a complete structured interview with follow-ups and review, use AI Mock Interview. The web assistant is better for quick validation and lightweight practice.