Overview

This guide covers AI transcription problems: not starting, poor accuracy, speaker detection issues, performance problems, and export errors.

Transcription Not Starting

1. Verify Feature is Enabled

   Settings → Features → Transcription

   • Toggle should be ON (blue)
   • If OFF, toggle it on
   • Wait 3 seconds for feature to activate

2. Check AI Model Status

   The AI model must be downloaded before transcription works:

   1. Settings → Features → Transcription
   2. Look for "Model Status" indicator
   3. If "Downloading..." - wait for download to complete
   4. If "Ready" - model is ready (green checkmark)
   5. If "Failed" - try "Retry Download" button

   First download is ~500MB. On slow internet, can take 5-15 minutes. Use wired connection if possible.

3. Check Audio Input is Active

   Transcription needs audio to transcribe:

   • Open Audio Routing panel (Ctrl + Shift + R)
   • Look for microphone or audio source with active level meter
   • Speak into mic - you should see green bar move
   • If no movement, check microphone (see Mic Troubleshooting)

4. Ensure Sufficient Free Disk Space

   Transcription needs ~100MB free space:

   • Right-click Drive C: → Properties
   • Check "Free space" - should be >1GB
   • If <500MB, free up disk space and restart Jyv

5. Check Transcription Permissions

   Windows may block transcription directory access:

   1. Settings → Privacy & security → Microphone
   2. Ensure Jyv Desktop is in allowed apps list
   3. Toggle OFF then ON to refresh

6. Restart Transcription Engine

   Sometimes the service gets stuck:

   • Settings → Advanced → Restart Transcription Engine
   • Wait 5 seconds for restart
   • Try transcribing again

7. Restart Jyv Desktop Completely

   Close Jyv (system tray icon → Exit) and reopen it.

Poor Accuracy

Transcripts have many errors or miss words.

Speakers Not Detected

1. Ensure Speaker Detection is Enabled

   Settings → Transcription → Identify Speakers

   • Toggle should be ON
   • If OFF, enable it for next transcription

2. Train Your Voiceprint

   Jyv learns to recognize your voice:

   1. Settings → Transcription → Speaker Training
   2. Click "Record Voice Sample"
   3. Speak naturally for 10-15 seconds
   4. Repeat 2-3 times for better recognition

3. Use Distinct Voices

   Speaker detection works best when voices are different:

   • Very similar voices confuse the AI
   • Different gender combinations work well
   • Different accents help
   • Can manually label speakers if detection fails

4. Check Audio Quality

   Poor audio hurts speaker detection:

   • Heavy echo reduces detection accuracy
   • Background noise interferes with voice identification
   • Enable Noise Suppression first

5. Increase Confidence Threshold

   Being too strict with speaker detection:

   • Settings → Transcription → Speaker Confidence
   • Lower threshold = more speakers detected (but more false positives)
   • Try "Medium" instead of "High"

6. Manually Label Speakers

   After transcription, manually add speaker labels:

   • Open transcript
   • Click speaker section → Edit label
   • Name the speaker
   • Jyv remembers for future transcripts

Slow Processing

1. Check CPU Usage

   Transcription is CPU-intensive. High CPU = slow processing:

   • Open Task Manager (Ctrl + Shift + Esc)
   • Check if other apps are using CPU
   • Close heavy apps (Chrome, video editor, games)

2. Switch to Fast Mode

   Trade accuracy for speed:

   • Settings → Transcription → Quality: Fast
   • Processes much faster, slightly less accurate

3. Disable Unused Features

   • Disable Speaker Detection temporarily
   • Disable Voice Enhancement
   • Disable Noise Suppression (if not needed)

4. Reduce Transcription Concurrent Sessions

   Only transcribe one source at a time:

   • Don't transcribe mic + Zoom + browser simultaneously
   • Focus transcription on one source

5. Restart Jyv Desktop

   Memory leaks can slow processing over time:

   • Exit Jyv completely
   • Reopen Jyv Desktop
   • Check if processing is faster

6. Check Disk Space

   Low disk space slows transcription:

   • Right-click Drive C: → Properties
   • If <1GB free, free up space
   • Jyv needs working space during processing
Processing Speed Expectations:
• Fast mode: Near real-time (1:1 ratio) with good CPU
• Balanced: Slightly slower than real-time on mid-range CPU
• Accurate: 0.5-1x real-time depending on CPU
• On low-end CPUs: can be 0.2-0.5x real-time

Crashes / Errors

Sync & Export Issues

Diagnostics

{
  "minimumRequirements": {
    "cpu": "Quad-core 2.4GHz+",
    "cores": 4,
    "ram": "8GB minimum, 16GB recommended",
    "disk": "500MB for model + 100MB working space",
    "ssd": "Recommended (faster processing)",
    "network": "500Mbps for initial 500MB model download"
  },
  "yourSystem": {
    "cpu": "Check Device Manager",
    "cores": "Task Manager -> Performance",
    "ram": "Settings -> System -> About",
    "freeSpace": "Right-click C: -> Properties"
  }
}