AI Interview Assistant (Desktop)

Start with the toolbar

The desktop client provides two input paths that can be used in the same practice session:

Screenshot solving organizes the answer for the task. Capture errors, test results or additional conditions to continue the same task. Use Start over or Reset view for a different question.

In the voice panel, Answer question generates a spoken interview answer; Voice solve handles tasks that need reasoning, a plan or an implementation. Concise answer affects Answer question only. Voice solving keeps the output structure required by its task.

First session

1. Complete installation and login and check permissions.
2. Choose your answer language, programming language and screenshot-solving model in Settings.
3. For image-based tasks, capture and check the queue before selecting Solve. For voice practice, open toolbar gear → Voice Settings, choose your role and audio setup, and wait for saving to finish. Click Voice or use the default Ctrl / ⌘ + M to start recording and open the panel, then choose Manual or Auto.
4. Check a real transcript and answer, then review the local or web record after stopping.

Computer call includes both computer speakers and headphones. Phone speaker uses the computer microphone to capture both speakers. Configure your role and audio setup in Voice Settings before recording; choose Manual or Auto in the voice panel after recording starts. See Voice assistant.

Reading and layout

• The toolbar keeps Screenshot and Solve; voice answering actions and Voice settings sit near the answer area.
• Settings → Advanced Settings → Content area size adjusts width and height separately. 100% restores the original size without changing text size. Use zoom shortcuts to scale text and controls together.
• Voice-answer shortcut scrolling uses a proportion of the visible area, capped at one screen per step as the height changes.
• The latest real conversation background is limited to 2,000 characters. The current answer target and retrieved reference material are handled separately.

Linking and translation

• Link: available to Pro and Ultimate. Pair a phone to view answers and control the desktop; choose whether to sync transcripts.
• Real-time translation: available to Ultimate. Supports Off, Auto, Chinese to English and English to Chinese. Translations are for display and do not enter the conversation background used for answers.

Local and web records

Local records prefer Documents/GankInterview Records/<session-time>/, including the Chinese-named interview and exam folders and model answer records. If Documents is unavailable, the client tries its application data directory.

Transcript lines are saved during recording and organized as Markdown when it ends. Solving records include answers and related captures; follow-up answers stay with the same task. Local copies do not depend on web upload success. Recordings of at least 60 seconds after rounding duration up to whole seconds with valid content are submitted for upload; queued records retry when the original account is signed in and the network is available.