Choose screenshot Q&A, live transcription Q&A, and knowledge-base workflows by scenario and run one clean interview loop end to end
Core Features
Remember this first
If you see a prompt, start with screenshot Q&A. If you are in a live conversation, start with live transcription Q&A. If you need answers to sound like your real background, prepare the Knowledge Base first and open the reading panel when you want visual support.
This page is about workflow, not settings. If you still need your baseline setup, start with Basic Settings. The default setup is already the recommended baseline for most interviews, so it is best to run one round as-is before deep tuning.
Pick the workflow by scenario
Screenshot Q&A
Best for online coding prompts, browser-based questions, algorithms, and take-home style tasks. The core pattern is capture first, then answer inside the screenshot Q&A view.
Live Transcription Q&A
Best for real-time interviews, behavioral rounds, project walkthroughs, and follow-up-heavy conversations. The core pattern is listen first, then organize the answer in the live transcription Q&A view.
Knowledge Base
Best when you need project details, resume facts, and answers that sound closer to your actual experience. The Knowledge Base powers retrieval, and the reading panel can stay open as an extra reference layer.
Screenshot Q&A
Use this for timed prompts, coding challenges, prompt pages, and visual question content. It is a "capture the visual context first, then answer inside screenshot Q&A" workflow. This is the capability many users still call Screenshot mode in the client.
Capture the prompt first
Prioritize the prompt, examples, constraints, and code area. One strong screenshot is often better than many noisy ones.
Generate with Ctrl/Cmd + Enter
Start with a first-pass answer, then decide whether to refine, ask for a shorter version, or capture an additional area. If you enable Auto-submit after screenshot in advanced settings, the current screenshot can jump straight into the screenshot Q&A view and save you one extra hotkey action.
Clear old screenshots when the question changes
If you move to a new prompt, clear the queue first. Old context is one of the most common reasons answers start drifting.
- Capture the question, examples, constraints, and function signature first.
- For the first pass, ask for approach, complexity, and edge cases before asking for a full final answer.
- If the question requires a specific coding format, make sure your baseline setting matches
General,LeetCode, orACMbefore the round starts. - If you will code manually, use the second turn to request a cleaner, more speakable solution outline.
- Prefer complete prompt sections over random cropped fragments.
- On long pages, start with the most important region instead of trying to capture everything at once.
- After generation, quickly verify that constraints and edge conditions were understood correctly.
- Keep the current question context and continue asking for bug fixes, complexity trade-offs, or alternative solutions.
- If the answer starts drifting, clear the screenshot queue and restart instead of forcing recovery inside stale context.
- Use AI as a fast reasoning scaffold, not as a script to copy line by line.
Most-used shortcut group
Ctrl/Cmd + H to capture, Ctrl/Cmd + Enter to generate, and Ctrl/Cmd + L to clear the queue.
If answers feel off
Check whether you mixed in an old screenshot, captured the wrong area, or kept stale context from a previous question.
Live Transcription Q&A
Use this for real-time Q&A, interviewer follow-ups, project walkthroughs, behavioral rounds, and system design conversations. It is a "listen first, transcribe live, then answer inside live transcription Q&A" workflow. This is the main workflow behind what users still see as Voice in the client.
Turn voice on when the conversation begins
Start Ctrl/Cmd + M a little before the actual exchange if possible, so the transcription loop can settle. Smart route is the safest default when you have not pinned a preferred route yet.
Use AI as structure, not a script
Live transcription Q&A works best when it gives you a clean answer skeleton that you rephrase naturally in your own voice.
Combine with screenshots when the discussion becomes visual
Use screenshot for diagrams, code, or whiteboard-style prompts while voice continues to carry the live conversation. The Knowledge Base reading panel can stay open as a third reference layer when needed.
- Voice mode is especially useful for "why this design", "what trade-off", and "what if production breaks" style follow-ups.
- If transcription quality dips, pause and recover instead of stacking more broken context.
- For long interviewer questions, wait until the full thought is complete before responding.
- If the round is mostly interviewer-led,
Interviewer onlyfiltering usually makes auto-triggered Q&A more focused; multi-speaker or discussion-heavy rounds often benefit from manual timing instead.
Ctrl/Cmd + 1switches to the live transcription Q&A view, which is best when you want real-time AI support while listening.Ctrl/Cmd + 2switches to the screenshot Q&A view, which is better when you need to keep visual prompts, code, or captured context in view.- If the round is both spoken and visual, keep the window open and switch between
1and2instead of restarting your flow.
- A short "context - action - result - reflection" structure usually works better than a long monologue.
- If your cloud resume, project notes, or Q&A docs are already in the Knowledge Base, retrieval plus the reading panel can keep answers much closer to your real projects.
- Compress the AI output into your own phrasing instead of delivering it like a script.
Knowledge Base Notes
Imported materials make answers much closer to your real projects, resume, and preferred language. The Knowledge Base note panel does not replace screenshot Q&A or live transcription Q&A. It works alongside them as a reference layer.
Usage reminder
The Knowledge Base note panel is stealth-friendly too. During a real interview, you can keep it open as a reference layer while continuing to use live transcription Q&A and screenshot Q&A. You do not have to choose only one.
Best materials to import first
Cloud resume, role-specific resume variants, project summaries, technical review notes, and behavioral story material.
Fastest way to see improvement
Documents with clean structure and strong details usually improve answer quality faster than simply uploading more files.
How it works alongside other modes
Keep the note panel open in stealth mode and let it supply project background, resume facts, and behavioral examples while Voice and Screenshot continue handling the active Q&A flow.
The most common interview loop
Prepare baseline settings
Lock language, model, permissions, devices, and shortcuts before the interview starts.
Pick screenshot or voice by the current moment
Use screenshot Q&A for prompt-first tasks and live transcription Q&A for live conversation. Combine them when the round becomes both visual and interactive.
Pull in Knowledge Base when personalization matters
This matters most in project walkthroughs, experience-heavy questions, and "tell me about a time when..." follow-ups. The Knowledge Base note panel can stay open without replacing the voice or screenshot workflow.
Rephrase and deliver naturally
The best workflow is not "copy AI". It is "use AI to think faster, then answer in your own style."
Continue reading
AI Reading Assistant
Send to your preferred AI
Smart Summary
Deep Analysis
Key Topics
Insights