Interview Training

The Interview Training feature is designed for actual interview scenarios, recognizing interviewer questions in real-time, retrieving relevant answers from the knowledge base, and providing professional answer suggestions. This feature helps you get real-time answer references during actual interviews to improve your interview performance.

Use Cases:

• Real Person Simulation: Through video tools like Tencent Meeting or Zoom, friends play the interviewer role and ask questions, the system automatically captures speaker audio
• Voice Simulation: Enter question text on translation websites (like Baidu or Google Translate), click the voice playback button, the system recognizes the question through speaker audio

1. Page Navigation

1.1 Prerequisites

Search for the CueMate application icon on the launcher and left-click to open the application.

After launching the CueMate desktop application, you can see the control window floating directly above the screen. The button functions of the control window from left to right are:

• "Visit CueMate website" icon: click to open the CueMate official website
• "Voice Recognition" button: Click to start the voice recognition function and open the voice function window
• "Ask AI" button: Click to open the AI question window, quickly ask AI questions and get answers
• "Interactive mode" icon: Switch the click penetration mode of the window (interactive mode can operate the window normally, clicks in penetration mode will penetrate to the underlying application)
• "Main application window" icon: Click to open the main application window and enter the background management interface
• "Show/Hide" icon: Click to hide or show the control bar and all other windows

1.2 Option 1: Launch from Control Bar

Steps:

1. Click the "Voice Recognition" icon on the floating control bar to open the voice recognition window.

2. In the popup window, click the "Interview Training" card.

Window Layout:

The system automatically opens three windows, displayed at the top center of the screen:

• Left Window: Interviewer window (displays interviewer questions and control panel)
• Middle Window: Interview conversation window (displays AI-generated reference answers and your real-time responses)
• Right Window: History window (records all completed Q&As)

1.3 Option 2: Launch from Home

Steps:

1. Click the "Main Application Window" icon on the floating control bar to open the main application window.

2. After selecting a position on the Home page, click the "LIVE Interview Training" button.

The system automatically opens the interview training window with the same layout as launching from the control bar, and automatically fills in your selected position information.

2. Start Interview Training

2.1 Confirm Training Configuration

In the left interviewer window, confirm the following settings:

Position Information

The top displays the current training position information card, including:

• Position name (including position description and resume JD information)
• Selected LLM (for generating answer suggestions)
• Voice language (Simplified Chinese/Traditional Chinese/English)
• Training mode (Manual mode/Auto mode, see Submit Answer)
  • Manual Mode (Default): You control the pace entirely. After interviewer asks a question, click "Question Complete" button to get AI answer. Suitable for friend mock interviews and repeated practice
  • Auto Mode: No manual operation needed. System automatically detects pauses and gets AI answer. Suitable for real online interview scenarios (Note: Long pauses may trigger falsely)

Click the expand button below the card to view complete position description and resume JD information.

Audio Device Selection

Configure audio devices:

• Microphone Device: Select microphone for recording your answers (optional)

  • System automatically lists all available audio input devices
  • Defaults to microphone configured in Voice Settings - Microphone Settings
  • This device records your answers for later review

• Speaker Device: Select speaker device for capturing interviewer voice (required)

  • System automatically lists all available audio output devices
  • Defaults to speaker configured in Voice Settings - Speaker Settings
  • Captures audio output from interview software (Tencent Meeting/Zoom) via AudioTee technology

IMPORTANT

• System Audio Capture Permission: Interview training requires "Screen Recording" permission to capture audio output from interview software. If not authorized, refer to permission authorization steps in Voice Test.

2.2 Start Training

After confirming configuration, click the "Start LIVE Training" button at the bottom of the left window.

System automatically executes the following initialization process:

1. Create Training Session

   • Create new interview training record with unique ID
   • Save position information, resume content, model configuration, etc.
   • Set training type to training

2. Initialize Context Management

   • Build resume and JD summary
   • Prepare conversation history management
   • Optimize context to reduce token consumption

3. Start Speaker Monitoring

   • Start AudioTee system audio capture
   • Connect to cuemate-asr speech recognition service
   • Begin monitoring audio output from interview software

4. Enter Listening State

   • Left window displays "Listening for interviewer questions..."
   • System ready, waiting for interviewer to start speaking

NOTE

AudioTee System Audio Capture: Through AudioTee technology, the system can capture audio output from interview software (Tencent Meeting, Zoom, DingTalk, Feishu, etc.) without interviewer cooperation, silently recognizing interviewer questions.

3. Training Process

3.1 Interviewer Asking Questions

When the interviewer starts speaking, the system automatically captures and recognizes the interviewer's voice.

Audio Source Methods

Interview training supports multiple interviewer voice sources:

1. Real Interview: Real interviews through video conferencing software like Tencent Meeting, Zoom, DingTalk, Feishu
2. Friend Simulation: Friend plays interviewer role through video conferencing software
3. Voice Simulation: Use translation website voice playback to simulate interviewer questions

Voice Simulation Method

If no real interviewer, use translation websites to simulate interviewer questions:

Open Baidu Translate, paste interview question text, click voice playback button:

Or use Google Translate:

Real-time Audio Capture

System captures interviewer voice through:

• AudioTee Capture: Captures audio output from interview software or translation website
• Real-time Transmission: Sends audio data to cuemate-asr speech recognition service in real-time
• Speech Recognition: cuemate-asr converts speech to text

Real-time Question Display

Recognized question text displays in real-time in the "Interviewer:" area at the bottom of the left window, updating as speech recognition progresses.

Confirm Question Complete

System cannot automatically determine where interviewer's question ends, need to confirm through following methods:

Method 1: Manual Mode (Default)

You control manually, more precise:

1. System captures and displays interviewer question in real-time
2. When you confirm interviewer finished asking, click "Question Complete" button in left window
3. System confirms question content, proceeds to next step

Method 2: Auto Mode

System auto-detects, more discreet:

1. System captures and displays interviewer question in real-time
2. System automatically detects when interviewer stops speaking
3. Automatically confirms question content, proceeds to next step

Mode Comparison

Feature	Manual Mode (Default)	Auto Mode
Confirmation	Click "Question Complete" button	System auto-detects
Precision	High (you control)	Medium (may misjudge)
Scenario	Interviewer speaks fast, often pauses	Interviewer speaks steadily, short questions
Advantage	Complete pace control, avoid misjudgment	No operation needed, more discreet
Disadvantage	Need to click button manually	Long pauses may misjudge

TIP

Mode Selection Suggestions:

• Manual Mode: Recommended for friend mock interviews, you can precisely control question boundaries
• Auto Mode: Recommended for real online interviews, reduce operations, avoid interviewer noticing

Recognition Quality Factors

• Volume: Interview software/translation website volume recommended at 50-80%
• Background Noise: System filters automatically, but excessive environmental noise affects recognition
• Speech Clarity: Clearer interviewer speech means more accurate recognition
• Network Latency: cuemate-asr service response speed (usually <1 second)

TIP

Optimize Recognition:

• Test speaker capture in Voice Test beforehand
• Adjust interview software/translation website volume appropriately (avoid too low or muted)
• Maintain quiet interview environment, reduce background noise

3.2 AI Reference Answer

After interviewer finishes asking (manual click "Question Complete" button or system auto-detects), AI generates reference answer.

1. Reference Answer Generation

Middle window starts generating reference answer:

• Reference answer displays in the center of middle window
• Content streams out word by word, you can prepare your answer while reading
• Answer generated based on your position JD, resume, question bank, etc.

2. History Update

Right history window synchronously adds a new Q&A card:

• Shows question number and interviewer's question content
• Shows AI-generated reference answer

3.3 Start Answering

After receiving AI reference answer, you can reference it when answering the interviewer. If microphone recording is enabled, system simultaneously records your answer.

Voice Recognition Area

Bottom of middle window shows voice recognition area:

• Recording Status Indicator: Circle on left shows ripple animation, indicating recording in progress
• Real-time Recognition Text: Right side shows real-time speech recognition results, initially displays "Waiting for voice input..."
• Auto Scroll: When recognized text is long, automatically scrolls to latest content

Current Mode Display

Right side of voice recognition display area shows current training mode (Manual/Auto):

• Mode selected before training starts, cannot switch during training
• Hover shows "Cannot switch mode during training"

Answer Complete Button

• In Manual Mode, when there's recognized text, "Answer Complete" button becomes enabled, click to submit answer
• In Auto Mode, "Answer Complete" button is disabled, system automatically submits after you stop speaking

TIP

Answering Tips:

• Quick Memory Points: Quickly scan key points of answer (2-3 core concepts)
• Natural Expression: Reorganize in your own words, don't read verbatim
• Combine Experience: Supplement with your own project experience details
• Maintain Communication: Maintain eye contact, don't look down at screen frequently
• Appropriate Pauses: Pause briefly after each point for more accurate speech recognition

3.4 Submit Answer

Interview training supports two submission modes, choose flexibly based on actual interview scenarios:

Mode Comparison

Feature	Manual Mode (Default)	Auto Mode
Scenario	Friend mock interview, repeated practice	Real online interview, smooth experience
Submission	Click "Answer Complete" button	System auto-detects
Advantage	Complete pace control, avoid false trigger	No operation needed, more discreet
Disadvantage	Need to click button manually	Long pauses may false trigger

Method 1: Manual Mode Submission

Workflow:

1. As you answer, system recognizes and displays your answer text in real-time
2. After finishing answer, click "Answer Complete" button (↵ icon)
3. System saves your answer, enters next round

Method 2: Auto Mode Submission

Workflow:

1. As you answer, system recognizes and displays your answer text in real-time
2. System auto-detects when you stop speaking
3. Auto submits answer, enters next round

TIP

Mode Selection Suggestions:

• Manual Mode: Recommended for friend mock interviews, can repeatedly practice, control pace
• Auto Mode: Recommended for real online interviews, reduce operations, avoid interviewer noticing

3.5 Enter Next Round

After submitting answer, system automatically enters next round.

Next Round Flow

1. Continue Listening

   • System re-enters listening for interviewer questions state
   • Left window displays "Listening for interviewer questions..."

2. Loop Process

   • Repeat "Interviewer Question → AI Reference Answer → Start Answering → Submit Answer" flow

View History

During interview training, you can view historical Q&A records anytime:

1. Click History Card

   • Click any Q&A card in right history window
   • Middle window switches to display that question's AI reference answer
   • Quick review of previous Q&A content

2. Review Purpose

   • Find points missed in previous answer
   • Review ambiguous previous answers, adjust later expression
   • Reference previous AI answers to supplement related knowledge points

3. Resume Current Conversation

   • Click restore button (🔄 icon) at bottom right of middle window
   • Immediately switch back to current question's AI reference answer
   • Continue current interview flow

4. Training Flow Control

During interview training, you can control training progress anytime, including pause, resume, and end training.

4.1 Pause Training

When to Pause

• Interviewer is introducing company background, team situation, etc. that doesn't need answering
• Interviewer leaves or temporarily handles affairs
• Need short break or drink
• Need to temporarily handle other matters

How to Pause

Two ways to pause training:

Method 1: Click "Pause" button (⏸️ icon) at bottom of left window

Method 2: Click "Pause" button (⏸️ icon) on top floating control bar (shown during training). Use expand/collapse button next to pause button to open or close interview training window.

After Pausing

• Speaker audio capture and speech recognition stop immediately
• Left window displays "Interview Training Paused"
• Current progress auto-saved, can continue after closing and reopening window

What You Can Do While Paused

• View all Q&A records in right history window
• Click history cards to view complete AI reference answers in middle window
• Think about and review generated answer key points

4.2 Resume Training

How to Resume

Two ways to resume training:

Method 1: Click "Resume" button (▶️ icon) at bottom of left window

Method 2: Click "Resume" button (▶️ icon) on top floating control bar (shown when training paused)

After Resuming

• System continues from where it paused
• Restarts speaker audio capture and speech recognition
• Left window displays "Listening for interviewer questions..."
• All historical Q&A records remain unchanged

4.3 End Training

When to End

• Interview completely finished
• Interviewer explicitly says interview is over
• Want to end current training session early

How to End

Two ways to end training:

Method 1: Click "Stop" button (⏹️ icon) at bottom of left window

Method 2: First pause training, then click "Complete" button (⏹️ icon) on top floating control bar (shown when training paused)

After Ending

• All Q&A records saved
• Left window displays "Interview Training Ended"
• Can continue viewing Q&A content in history

4.4 View Interview Review Records

After training ends, click "Interview Review" in top menu to enter Interview Review page.

Interview Review page displays all historical training records in timeline format, click "View Details" to see complete training report.

NOTE

Report Generation Conditions: If you recorded your answers during training, system generates evaluation report. Only Q&A records with candidate answers recorded will be analyzed and scored.

For detailed feature description, see Interview Review documentation.

5. Feature Description

5.1 How It Works

Complete interview training flow:

Start Training
  ↓
[Capture Audio] → System Audio Capture (AudioTee)
  ↓
[Recognize Question] → Speech Recognition (cuemate-asr) → Real-time Text
  ↓
[Trigger Generation] → Manual Click / System Auto-detect
  ↓
[Search Question Bank] → RAG Semantic Search (similarity ≥0.8)
  ↓
  ├─ Match Found → Return question bank answer
  └─ No Match → AI real-time generation (based on JD + Resume + Question)
  ↓
[Display Answer] → Stream display in right window
  ↓
Loop until training ends

5.2 Question Bank Matching Mechanism

System matches questions through following steps:

1. Generate Question Vector: Convert interviewer question to embedding vector
2. Semantic Search: Search position's questions in ChromaDB
3. Dual Filtering:
   • Vector similarity ≥0.8
   • Keyword hit rate ≥80%
4. Return Best Match: Get 1 question with highest similarity and its answer

Example:

• Interviewer asks: "Please explain Redis persistence mechanisms"
• Question bank has: "What are Redis persistence methods?"
• Match successful, return question bank answer

5.3 System Audio Capture Technology

Interview training uses AudioTee technology for system audio capture:

• How It Works: Captures audio output from interview software (Tencent Meeting/Zoom)
• Advantage: No interviewer cooperation needed, silent capture
• Permission Required: "Screen Recording" permission needed

Supported Interview Software:

• Tencent Meeting
• Zoom
• DingTalk
• Feishu
• Any other audio/video software

6. Notes

6.1 Prerequisites

Before starting interview training, ensure completed:

1. Create Position (Required) - Create interview position, fill in JD and resume
2. Model Settings (Required) - Configure LLM
3. Voice Settings (Required) - Configure speaker device, test system audio capture
4. Interview Questions (Strongly Recommended) - Preset common interview questions for better answer accuracy
5. Vector Knowledge Base Sync (Strongly Recommended) - Ensure questions synced to ChromaDB
6. Voice Test (Recommended) - Test speaker capture capability

6.2 Service Dependencies

Interview training depends on following services:

• cuemate-asr Service: Speech recognition (port 10095)
• LLM Router Service: LLM routing (port 3002)
• RAG Service: Question bank retrieval (port 3003)
• Web API Service: Data management (port 3001)
• ChromaDB Service: Vector database (port 8000)

6.3 Permission Requirements

• Screen Recording Permission: For system audio capture (required)
  • Path: System Preferences > Security & Privacy > Screen Recording > Allow CueMate
• Microphone Permission: For recording your answers (optional)

6.4 Performance Optimization

1. Question Bank Preparation:

   • Add common questions in Interview Questions beforehand
   • Ensure synced in Vector Knowledge Base
   • Question answers should be detailed and accurate

2. Model Selection:

   • Recommend GPT-3.5-turbo or fast domestic models
   • Ensure fast response (❤️ seconds) to avoid interviewer waiting

3. Network Requirements:

   • Ensure stable network to avoid API timeout
   • Using domestic models can improve response speed

4. Audio Devices:

   • Adjust interview software volume appropriately
   • Avoid volume too low causing recognition failure

7. Best Practices

7.1 Training Preparation

1. Thoroughly Prepare Question Bank:

   • Cover position-related technical points
   • Include common behavioral interview questions
   • Answers should be detailed, structured, contain keywords

2. Environment Preparation:

   • Quiet room, avoid environmental noise
   • Test system audio capture to ensure normal
   • Ensure stable network to avoid API timeout

3. Device Preparation:

   • Confirm screen recording permission granted
   • Adjust interview software volume appropriately
   • Prepare paper and pen for noting answer key points

7.2 Usage Tips

1. Reference, Don't Read:

   • AI answers are reference suggestions, don't read verbatim
   • Reorganize in your own words
   • Supplement with your own project experience

2. Quick Memory Points:

   • After answer appears, quickly scan key points
   • Remember 2-3 core concepts
   • Express in your own words

3. Flexible Response:

   • Adjust answer based on actual situation
   • Don't use directly if doesn't match resume
   • When interviewer follows up, reference previous answers

4. Natural Communication:

   • Don't frequently look down at screen
   • Maintain eye contact (camera)
   • Natural speaking pace, don't sound like reciting

7.3 Mode Selection Suggestions

Scenario	Recommended Mode	Reason
Friend Mock Interview	Manual Mode	Easy to control pace, repeated practice
Real Online Interview	Auto Mode	Reduce operations, avoid interviewer noticing
Fast-speaking Interviewer	Manual Mode	Avoid false trigger, ensure complete question
Slow-speaking Interviewer	Auto Mode	Auto-detect, save time
Unstable Network	Manual Mode	Avoid auto-trigger then API timeout

7.4 Risk Notice

When using interview training, please note:

1. This is an assistive tool, don't rely on it completely
2. Be honest during actual interviews
3. Answers are for reference only, combine with actual situation
4. Don't read mechanically, express naturally
5. Some companies may prohibit assistive tools, judge for yourself

8. FAQ

8.1 Cannot Capture Interviewer Voice

Problem: Left panel not showing interviewer speech text

Solutions:

1. Check if "Screen Recording" permission granted to CueMate
2. Confirm speaker device selected correctly
3. Adjust interview software volume (recommended 50-80%)
4. First do Voice Test to confirm speaker capture works
5. Check error messages in Log Management

8.2 Question Bank Not Matching

Problem: AI always generates answers in real-time, never uses question bank

Solutions:

1. Check if questions added in Interview Questions
2. Confirm questions synced in Vector Knowledge Base
3. Check if question's position matches current training position
4. Try adding more related questions to increase match probability

8.3 Auto Mode False Triggers

Problem: In auto mode, answer generation triggers before interviewer finishes

Solutions:

1. Switch to manual mode, you control trigger timing
2. Wait for interviewer to clearly finish before confirming question
3. Check if environmental noise being misrecognized as interviewer voice

8.4 AI Answers Inaccurate

Problem: AI-generated answers not relevant to question

Solutions:

1. Check if position JD and resume filled completely
2. Confirm interviewer question recognized accurately (text in left panel)
3. Switch to higher quality LLM (e.g., GPT-4)
4. Add similar questions to question bank to improve match rate

8.5 Answer Generation Delay

Problem: AI answer generation very slow

Solutions:

1. Switch to faster responding model (GPT-3.5-turbo, Zhipu, DeepSeek)
2. Check if network connection stable
3. Use domestic models to reduce network latency
4. Prepare question bank in advance (question bank returns faster)

9. Feature Comparison

9.1 Mock Interview vs Interview Training

Comparison	Mock Interview	Interview Training
Scenario	Solo practice, advance preparation	Actual interview, real-time assistance
Interviewer	AI plays role	Real interviewer
Question Method	AI actively asks	Interviewer asks, system captures
Answer Method	AI gives suggestions, you voice answer	AI gives suggestions, you reference answer
Training Effect	Familiarize process, improve adaptability	Get answer reference, improve accuracy
Audio Capture	Microphone, Speaker	Microphone, Speaker

Related Pages

• Home - Start interview training
• Create Position - Create interview position
• Interview Questions - Prepare interview questions
• Vector Knowledge Base - Sync questions to vector database
• Interview Review - View training records
• Mock Interview - AI mock interview training
• Voice Test - Test audio devices
• Voice Settings - Configure speech recognition
• Model Settings - Configure LLM