Copilot cho phỏng vấn và họp của bạn.

What is Quiet Assistant?

Quiet Assistant is an AI-powered live copilot for interviews and meetings. It transcribes the other side's voice from your system audio in real time and generates Claude-powered answers when they pause, personalized with the user's resume and the role's job description. For meetings where AI replies on every pause would be noisy, an auto-trigger toggle (⌘⇧L) turns it into a quiet transcription tool with answers fired only on demand (⌘⇧⏎). It also analyzes screenshots of any problem on screen, auto-classifying them as coding (full code + complexity) or general (direct paragraph answer). The user's own microphone is not captured in production builds, keeping their side of the conversation private.

Key Features:

• Live Audio Sessions (⌘L, Pro) — continuous transcription of the other side's voice, auto-generated answers
• Quiet transcription mode (⌘⇧L off) for meetings — transcript without continuous AI replies
• Trigger Now (⌘⇧⏎) — fire an AI answer manually only when wanted
• Resume + JD personalized answers grounded in the user's real experience
• Behavioral / "tell me about a time" answers structured in STAR format (Situation / Task / Action / Result) with a likely follow-up question
• Local Session History (⌘⇧H toggle) — plain JSON on your device, never uploaded; full-text search across past sessions, export to Markdown or JSON
• Auto-classified screenshot analysis — coding vs. general, no mode toggle
• Multi-monitor screenshot — both ⌘S and ⌘⇧S follow the focused window's display, so the overlay always lands on the screen you're looking at
• Input language lock for live ASR (⌘⌥←/→) — pin one of 16 supported languages so heavy accents don't flip mid-session; response language cycles independently with ⌘⇧←/→
• Inline speaker-side translation across 16 languages, fired only when the input language differs from the response language
• Production builds don't capture the user's mic — only the other side is transcribed
• Multi-platform: macOS 13+ via ScreenCaptureKit, Windows 10+ via Chromium WASAPI loopback

How does Quiet Assistant work?

Quiet Assistant supports two complementary workflows in a single desktop app:

Live Audio workflow (interviews and meetings):

1. Press ⌘L (Pro) to start a Live Audio Session
2. The app captures the other side's voice from your system audio and shows a live transcript on the panel
3. For interviews: when they pause, Claude generates a 3–5 sentence answer using your saved resume and JD
4. For meetings: toggle auto-trigger off (⌘⇧L) — the panel becomes a quiet transcript tool; press ⌘⇧⏎ to fire an answer only when you actually want one
5. On session end, the full transcript and answers are auto-saved as plain JSON in your app's userData directory — never uploaded; open with ⌘⇧H
6. Your own microphone is not captured in production builds, so what you say never leaves your room

Screenshot workflow (visible problems):

1. Press ⌘S (full screen) or ⌘⇧S (area selection) to capture the problem — the capture follows your focused window's monitor on multi-display setups
2. Press ⌘Return to analyze — Claude auto-classifies the screenshot
3. Coding problems get full code, complexity analysis, and follow-up Q&A
4. General questions (system design, math, concept) get a direct paragraph answer with key points
5. Behavioral / "tell me about a time" prompts arrive in STAR format (Situation / Task / Action / Result) on separate lines, plus one likely follow-up question

Can I use Quiet Assistant during meetings?

Yes. Press ⌘L to start a Live Audio Session, then toggle auto-trigger off with ⌘⇧L. The transcript runs continuously on the side panel without firing AI replies on every pause. Press ⌘⇧⏎ ("Trigger Now") any time you want an AI answer on the recent transcript. Designed for stand-ups, design reviews, sales calls, customer interviews, and any meeting where you want a transcript and an occasional assist rather than continuous AI suggestions. The transcript is stored locally as plain JSON in your app's userData directory (⌘⇧H to review) — never uploaded; FileVault / BitLocker covers data-at-rest. Your own mic is never captured in production builds.

Quiet Assistant vs Other Interview Assistants

Quiet Assistant combines the live-audio answering of LockedIn AI / Final Round AI / InterviewCopilot.io with resume + JD personalization (a Beyz AI pattern), plus a quiet-transcription mode that makes it usable for ordinary meetings — not just interviews.

Differentiators:

• Works for both interviews (continuous AI replies) and meetings (quiet transcription, AI on demand)
• Live Audio Sessions on both macOS (ScreenCaptureKit) and Windows (Chromium WASAPI loopback)
• Resume + JD personalization shared between Live Audio answers and screenshot analysis
• Local-only session history with full-text search and Markdown / JSON export — plain JSON on your device, never uploaded
• Input language lock for live ASR (16 languages) — pin one to keep heavy accents from flipping mid-session
• Inline speaker-side translation that only fires when the input language differs from the response language
• Flexible pricing: $10/week, $30/month, or $180/year ($15/month billed annually)
• Single desktop app — no browser extension required
• Powered by Claude Sonnet 4.5 with Haiku for translation and lightweight tasks

Quiet Assistant Pricing

Quiet Assistant offers both free and paid plans:

Free Plan:

• Auto-classified screenshot analysis with blurred results
• Test the complete workflow before subscribing
• No Live Audio Sessions on the free tier

Pro Plans:

• $10/week — fits short interview loops
• $30/month — standard monthly billing
• $180/year — works out to $15/month billed annually
• 3-day free trial on every Pro plan
• Live Audio Sessions for interviews and meetings, resume + JD personalization, quiet-transcription mode, local session history, inline translation
• 16 input + 16 response languages
• Unmetered usage within the plan
• Email support

How to Install Quiet Assistant

Installing Quiet Assistant is simple and takes less than 2 minutes:

Installation Steps:

1. Download the installer from quietassistant.io
2. Run the installer (no administrator rights needed)
3. Complete the one-time Google authentication
4. Grant Screen Recording permission so the app can capture system audio (the other side's voice) for Live Audio Sessions
5. Save your resume and JD on the dashboard for personalized answers
6. Press ⌘L to start a Live Audio Session, or ⌘S to capture a screenshot

System Requirements:

• macOS 13+ (Ventura) for Live Audio Sessions; macOS 10.15+ for the screenshot workflow only
• Windows 10/11 for full Live Audio Sessions
• 4GB RAM minimum
• Internet connection
• 500MB free disk space

Best AI Interview and Meeting Copilot 2026

Quiet Assistant is widely picked as a leading AI live-audio copilot in 2026 because it combines live answering for interviews, quiet transcription for meetings, resume-personalized responses — all in a single desktop app.

Why people pick Quiet Assistant in 2026:

• Live Audio Sessions answer interviewer questions automatically — works on macOS and Windows
• Quiet transcription mode (⌘⇧L off) makes the same tool usable for stand-ups, design reviews, and sales calls
• Answers grounded in the user's real resume and the role's job description
• Local-only session history (plain JSON) — never uploaded
• Inline translation only when the other side's language differs from the user's target
• Flexible pricing: $10/week, $30/month, or $180/year for short or long usage
• Screenshot workflow works alongside HackerRank, CodeSignal, CoderPad, LeetCode OA, and any browser-based platform

What AI model does Quiet Assistant use?

Quiet Assistant is powered primarily by Claude Sonnet 4.5 (Anthropic) for answer generation in both Live Audio Sessions and screenshot analysis. Claude Haiku 4.5 handles lightweight inline translation. Audio transcription uses OpenAI's gpt-4o-transcribe (chunked streaming, sub-second latency). Claude is known for superior reasoning and code generation — ideal for technical interview answers under time pressure.

Quiet Assistant pricing and plans 2026

Quiet Assistant offers a free trial and three Pro tiers: $10 for a week, $30 for a month, or $180 for a year (works out to $15/month when billed annually). All Pro plans include Live Audio Sessions for interviews and meetings, resume + JD personalization, quiet-transcription mode, local session history (plain JSON, never uploaded), inline speaker translation, 16 input + 16 response languages, and unlimited usage.

How to prepare for Meta's AI-enabled coding interview?

Meta's AI-enabled coding interview (introduced in 2025) allows candidates to use AI tools like Claude Sonnet and GPT-4o during a 60-minute coding session in CoderPad. To prepare: practice using AI tools to solve algorithmic problems, learn to verify and explain AI-generated solutions, work on multi-file project structures, and practice time management across problem exploration, design, implementation, and optimization phases. Quiet Assistant's auto-classified screenshot analysis and Live Audio Sessions are ideal for simulating this format — capture the CoderPad problem with ⌘S for code generation, then use ⌘L during the design discussion.

What are Live Audio Sessions in Quiet Assistant?

Live Audio Sessions are Quiet Assistant's flagship Pro feature for live conversations — interviews, behavioral rounds, and meetings. Pressing ⌘L starts continuous capture of the other side's voice via system audio (ScreenCaptureKit on macOS 13+, Chromium WASAPI loopback on Windows 10+). The transcript is shown live on the side panel. When they pause, Claude Sonnet 4.5 generates a concise spoken answer (3–5 sentences) personalized with the user's saved resume and the target job description. Behavioral / "tell me about a time" prompts arrive structured in STAR format (Situation / Task / Action / Result) with one likely follow-up. Auxiliary shortcuts: ⌘⇧L toggles auto-trigger on or off (turn it off for quiet meetings), ⌘⇧⏎ ("Trigger Now") fires an answer on demand, ⌘⇧W closes the panel, ⌘⇧H toggles the local session history window (with full-text search across all past sessions and Markdown / JSON export), ⌘⌥←/→ cycles the input language to lock the spoken language onto one of 16 supported languages (so heavy accents don't flip mid-session), and ⌘⇧←/→ cycles the AI's response language separately. The user's own microphone is not captured in production builds. Transcripts of the other side's voice are stored locally as plain JSON in the OS's standard application-data directory (`~/Library/Application Support/interview-assistant/sessions/` on macOS, `%APPDATA%\interview-assistant\sessions\` on Windows) — never uploaded; data-at-rest is covered by your OS-level disk encryption (FileVault / BitLocker).

How does Quiet Assistant handle behavioral interview questions?

When Claude Sonnet 4.5 classifies a screenshot or live-audio question as behavioral / situational ("tell me about a time when...", "describe a conflict you...", "how did you handle..."), the answer arrives structured in STAR format on separate lines: Situation: the context, Task: the candidate's responsibility, Action: what they did, Result: the outcome. The whole STAR block stays around 5–7 sentences so it's scannable mid-interview rather than a wall of prose. Below the answer, a Follow-up: field carries one likely next question so the candidate can pre-empt it. The personalization layer pulls from the user's saved resume, so the Situation and Action draw on real experience rather than generic templates.

Can I search and export Live Audio session history?

Yes. Press ⌘⇧H to toggle the Session History window — a second press closes it. The search bar above the session list filters across every session's full event body (interviewer transcripts, your own transcripts when present, inline translations, AI suggestions) and surfaces a small italic snippet with the matched substring highlighted. Typing debounces at 200ms; ESC clears the query first, then closes the window. Each session detail has Export Markdown / Export JSON buttons next to Delete: Markdown is human-readable with timestamps, role headings, and translations as block-quotes; JSON is the raw session record for piping into your own tooling. Sessions live as plain JSON in the app's userData directory — exports go through the OS file dialog and write to disk only; nothing is uploaded.