# Live Transcription

Display real-time captions on listener devices with on-device speech recognition.

## What is Live Transcription?

Live Transcription converts spoken audio into text captions that appear on listeners' devices in real-time. Perfect for:

- Attendees who are deaf or hard of hearing
- Non-native speakers following along
- Noisy environments where audio is difficult to hear
- Recording accurate notes or minutes
- Translation preparation

**Privacy First:** All speech recognition happens on your Mac using Apple's on-device technology. No audio is sent to cloud services.

---

## Enabling Live Transcription

### Step 1: Grant Permission

The first time you enable Live Transcription, macOS will ask for permission:

1. A system dialog will appear requesting Speech Recognition access
2. Click **"OK"** to grant permission
3. Audion needs this permission to convert audio to text

> **Note:** If you accidentally deny permission, go to **System Preferences** → **Security & Privacy** → **Privacy** → **Speech Recognition** and enable it for Audion.

### Step 2: Enable in Preferences

1. Open Audion
2. Go to **Preferences** (⌘,)
3. Click the **"Speech Recognition"** tab
4. Check **"Enable Speech Recognition"**
5. Select your preferred **language** from the dropdown

### Step 3: Start Streaming

1. Start your audio stream as normal
2. Transcription will begin automatically when audio is detected
3. Listeners will see captions appear on their devices in real-time

---

## Language Support

Live Transcription supports **40+ languages** including:

### English Variants
- English (United States)
- English (United Kingdom)
- English (Australia)
- English (Canada)

### Major Languages
- Spanish (US, Spain, Mexico)
- French (France, Canada)
- German
- Italian
- Japanese
- Korean
- Chinese (Simplified, Traditional - China, Taiwan, Hong Kong)
- Portuguese (Brazil, Portugal)
- Russian
- Arabic

### Additional Languages
Dutch, Swedish, Danish, Finnish, Norwegian, Polish, Turkish, Thai, Indonesian, Vietnamese, Hebrew, Romanian, Czech, Slovak, Ukrainian, Croatian, Catalan, Greek, Hungarian, Malay, Hindi

> **Tip:** For best accuracy, select the language variant that matches your speaker's accent and dialect.

---

## Text Corrections

Improve transcription accuracy with automatic text corrections.

### Biblical Terms Correction

Automatically corrects common speech recognition errors for religious terms:

**Enable Biblical Terms:**
1. Preferences → **Text Corrections** tab
2. Check **"Use Biblical Terms corrections"**

**What it corrects:**
- Bible book names (Genesis, Exodus, Revelation, etc.)
- Common religious terms (Jesus, Holy Spirit, Gospel, etc.)
- Scripture references (verse, chapter, Scripture, Bible)
- Religious ceremonies (Baptism, Communion, Eucharist, etc.)

**Examples:**
- "revelations" → "Revelation"
- "zachariah" → "Zechariah"
- "first corinthians" → "1 Corinthians"
- "holy spirit" → "Holy Spirit"

### Custom Corrections

Add your own text corrections for specialized vocabulary:

**Add a Custom Correction:**
1. Preferences → **Text Corrections** tab
2. Click the **"+"** button
3. Enter the **incorrect** text (what speech recognition says)
4. Enter the **correct** text (what it should be)
5. Click **"Add"**

**Common Uses:**
- Organization names ("saint mary's" → "St. Mary's")
- People's names ("john smith" → "Rev. John Smith")
- Technical terms ("data base" → "database")
- Brand names ("face book" → "Facebook")
- Local place names

**Managing Corrections:**
- **Edit:** Double-click any correction to modify it
- **Delete:** Select a correction and click the **"-"** button
- **Disable All:** Uncheck "Enable Text Corrections" to temporarily turn off

> **Tip:** Custom corrections override biblical terms, so you can create your own preferred spellings.

---

## Punctuation Enhancement

Audion automatically enhances punctuation for better readability:

**Automatic Features:**
- Capitalizes first letter of sentences
- Adds question marks for interrogative sentences
- Inserts commas after introductory phrases
- Adds periods to complete sentences
- Splits long run-on sentences

**Examples:**
- "what is your name" → "What is your name?"
- "however this is important" → "However, this is important."
- "yes we can help" → "Yes, we can help."

---

## How Listeners See Transcriptions

### On Web Devices

When listeners connect to your stream:

1. They see the standard audio player
2. Below the audio player, a **transcription area** appears
3. Text appears in real-time as you speak
4. Previous transcriptions remain visible (scrollable)
5. Final transcriptions appear with enhanced punctuation

**Display Features:**
- Large, readable text
- Smooth scrolling
- Automatic scroll-to-bottom for new text
- Language indicator
- Visual distinction between interim and final text

### Transcription Modes

- **Interim Text (gray):** Appears immediately as you speak, updates in real-time
- **Final Text (black):** Locked-in transcription with corrections and enhanced punctuation

---

## Best Practices

### For Clear Transcriptions

✅ **Do:**
- Speak clearly and at a moderate pace
- Use a good quality microphone
- Minimize background noise
- Speak in complete sentences when possible
- Allow brief pauses between thoughts

❌ **Avoid:**
- Speaking too quickly
- Mumbling or speaking very softly
- Overlapping speech (multiple people talking)
- Heavy background music during speech
- Sudden loud noises

### Audio Setup Tips

**Recommended Setup:**
- Use the same audio input for transcription and streaming
- Test transcription before your event
- Verify corrections are working properly
- Have a backup speaker/mic in case of technical issues

**Microphone Distance:**
- **Too close (< 6 inches):** May pick up breathing, pops
- **Optimal (6-12 inches):** Clear speech, good accuracy  
- **Too far (> 24 inches):** Reduced accuracy, more errors

---

## Troubleshooting

### "Speech Recognition Not Authorized"

**Solution:**
1. Go to **System Preferences** → **Security & Privacy**
2. Click **Privacy** tab
3. Select **Speech Recognition** from the list
4. Check the box next to **Audion**
5. Restart Audion

### No Transcriptions Appearing

**Check these:**
- ✓ Speech recognition is enabled in Preferences
- ✓ Correct language is selected
- ✓ Audio levels are showing in the meter
- ✓ Your Mac supports on-device recognition (macOS 11+)
- ✓ Streaming is active

### Transcriptions Are Inaccurate

**Try these fixes:**
1. **Select correct language variant** (e.g., US vs UK English)
2. **Enable text corrections** for your domain (biblical terms, custom)
3. **Improve audio quality:**
   - Get closer to microphone
   - Reduce background noise
   - Use a better quality mic
4. **Speak more clearly:**
   - Slow down slightly
   - Enunciate words
   - Pause between sentences

### Corrections Not Working

**Verify:**
1. "Enable Text Corrections" is checked in Preferences
2. Your corrections are entered correctly (case-insensitive)
3. Words match exactly what speech recognition produces
4. Try the transcription feature alone to see raw output

**Debug Process:**
1. Disable all corrections temporarily
2. Speak your test phrase
3. Note exactly what appears
4. Add that exact text as the "incorrect" value
5. Enable corrections and test again

### Transcriptions Lag Behind Audio

**Normal latency:** 100-300ms is expected

**If lag is excessive:**
- Check Mac CPU usage (Activity Monitor)
- Close unnecessary applications
- Reduce number of simultaneous listeners
- Ensure Mac meets system requirements

### Text Appears But Then Disappears

**This is normal behavior:**
- **Interim text (gray)** shows immediately but updates as recognition improves
- **Final text (black)** appears when recognition is confident
- Wait a moment for final transcription with corrections

---

## Performance Tips

### System Resources

**Speech recognition uses CPU power:**
- Modern Macs (M1/M2/M3): No issues
- Intel Macs: May see 10-20% CPU usage
- Older Macs: Test performance before live use

**Optimize Performance:**
- Close unused applications
- Disable other speech features (Siri, Dictation)
- Use wired network connection for Mac
- Ensure adequate ventilation to prevent thermal throttling

### Network Bandwidth

**Transcription adds minimal bandwidth:**
- Text data is very small (~1-5 KB per sentence)
- Negligible impact compared to audio streaming
- Works well even on congested networks

---

## Privacy & Security

### On-Device Processing

**Your audio stays private:**
- ✅ All speech recognition happens on your Mac
- ✅ No audio sent to Apple or third parties
- ✅ No cloud processing required
- ✅ Works without internet connection

### What's Stored

**Temporary only:**
- Transcriptions are generated in real-time
- Text is broadcast to listeners immediately
- Nothing is saved or recorded by Audion
- Listeners can't export or save transcriptions

**Listener Privacy:**
- Audion doesn't identify who receives transcriptions
- No personal data collected from listeners
- Same privacy as audio streaming

### Compliance Considerations

**Accessibility:**
- Live transcription helps meet ADA requirements
- Provides accommodation for deaf/hard of hearing
- Complements other assistive technologies

**Accuracy Disclaimer:**
- Machine transcription may contain errors
- Not a substitute for professional CART services
- Best for real-time assistance, not legal records

---

## Frequently Asked Questions

### Can I use transcription without streaming audio?

No, transcription requires an active audio stream. The same audio input is used for both features.

### Does transcription work with all languages simultaneously?

No, select one language at a time. For multilingual services, you would need to change the language setting between segments.

### Can listeners turn off transcriptions?

Not currently. Transcriptions appear for all listeners when enabled.

### Do transcriptions work offline?

Yes! On-device speech recognition works without internet. However, initial setup may require downloading language models (happens automatically).

### Why is accuracy better for some speakers?

**Factors affecting accuracy:**
- Speaker's clarity and pace
- Accent and dialect match to selected language
- Microphone quality and positioning
- Background noise levels
- Technical vocabulary vs. common words

### Can I export transcriptions?

Audion doesn't save or export transcriptions. For records, use separate screen recording or transcription software.

### Does this replace CART services?

No, this is real-time machine transcription. For legal proceedings, educational settings, or situations requiring high accuracy, use professional CART (Communication Access Realtime Translation) services.

### Which macOS version do I need?

**macOS 11 (Big Sur) or later** for basic recognition. macOS 13 (Ventura) or later provides enhanced punctuation.

---

## What's Next?

- **Customize branding:** Add your logo with [Customization](customization.md)
- **Automate control:** Set up [MIDI Control](midi-control.md)
- **Audio optimization:** Improve quality with [Audio Setup](audio-setup.md)
- **Need help?** See [Troubleshooting](troubleshooting.md)

---

**Have feedback about Live Transcription?** Email us: [me@josiaho.com](mailto:me@josiaho.com?subject=Re:Audion%20Live%20Transcription)