# OCvoice — Comprehensive Product Guide

> Real-time AI translation for churches. Dozens of languages (full list below), no app install, theological precision.

OCvoice by OCplan ApS is a real-time AI-powered translation platform purpose-built for churches and religious services. Congregants scan a QR code on their phone, choose their language, and hear the sermon translated live through AirPods or read subtitles — no app download required.

## Executive Overview

OCvoice solves the critical challenge of making sermons and religious content accessible to multilingual congregations without expensive interpretation equipment or coordination. A pastor speaks naturally in Danish (or Danish + English mixed-speech), and listeners instantly receive high-quality translations in any of the supported languages on their personal devices.

### Key Differentiators

- **Theological Precision**: Built-in fully stocked theological glossary — curated renderings in every supported language ensure theological meaning is never distorted
- **No App Required**: Progressive Web App (PWA) — listeners simply scan a QR code and start listening in seconds
- **Real-time Accuracy**: Intelligent AI routing ensures both speed and quality across all language complexities
- **Code-switching Support**: Native support for Danish-English mixed speech common in international churches
- **Enterprise-grade**: Real-time subscriptions, listener-aware translation gating, and speaker diarization for professional quality
- **Full Price Transparency**: Pay-as-you-go per-minute pricing with no long-term contracts

---

## Technical Stack

### Architecture Overview

```
Broadcaster Stream
  ↓
Advanced Speech Recognition (ASR)
  ├─ Real-time Danish/English code-switching
  ├─ Speaker diarization
  └─ Profanity filtering
  ↓
Serverless Backend
  ├─ Transcript ingestion + storage
  ├─ Intelligent AI translation routing
  ├─ High-quality TTS generation
  └─ Real-time subscriptions for listeners
  ↓
Listener Devices (Web Browser)
  ├─ Subtitles in chosen language
  └─ Synthesized audio (AirPods, speaker, etc.)
```

### Core Technologies

**Frontend**
- Modern web framework with server-side rendering
- Utility-first responsive design system with OCplan brand palette
- Type-safe frontend code
- Edge-deployed progressive web app

**Backend**
- Serverless, type-safe backend
- Real-time subscriptions, database, file storage, and scheduled jobs
- Fully managed real-time layer — no separate gateway or cache to operate

**AI/ML & Speech**
- **Advanced Speech Recognition API** — Real-time speech recognition with:
  - Danish and English native support with code-switching awareness
  - Speaker diarization (identifies different speakers)
  - Low latency recognition
  - Automatic profanity filtering

- **Specialized AI Models** — Intelligent multi-tier translation engine:
  - Fast, cost-effective models for strong/major languages
  - Maximum quality models for weak/low-resource languages
  - Prompt caching for system prompt reuse
  - Glossary-enforced system prompt (curated theological concepts)

- **High-Quality Text-to-Speech** — Multi-tier voice synthesis:
  - Premium voices for major languages
  - High-definition synthesis options
  - AI-powered voice generation for lower-resource languages
  - Multi-gender voice selection (male/female configurable per organization)

**Authentication**
- Enterprise SAML/SSO integration
- JWT-based role management (admin/member roles)

**Infrastructure**
- Monorepo package management
- Modern serverless database with type-safe schema validation
- Relational database architecture

### Core Modules

The backend includes specialized modules for:
- Real-time ASR transcript ingestion and profanity filtering
- Intelligent multi-tier AI translation routing
- Multi-tier text-to-speech voice synthesis
- Session management and broadcaster locking
- Billing and usage tracking
- Organization settings and TTS configuration
- Theological glossary management and lookup
- Speaker profiles and voice enrollment
- Post-session quality scoring

---

## Translation Quality & Routing Strategy

### Intelligent Model Routing

OCvoice uses intelligent model selection based on language linguistic complexity. The system automatically routes between:

- **Fast + Cost-effective Models**: Used for languages with abundant training data and well-established neural machine translation (most European languages, major Asian languages, Arabic, Hebrew, and others)

- **Maximum Quality Models**: Reserved for low-resource and challenging languages requiring maximum precision (endangered/regional languages, morphologically complex languages, and specialized linguistic structures)

Organizations can customize translation behavior:
- Automatic two-tier routing (default)
- Always use fast models (faster, cheaper)
- Always use maximum quality models (highest quality)
- Theological precision mode: Forces maximum quality + enriched glossary for maximum theological accuracy

### Glossary-Enforced System Prompt

All translations enforce a fully stocked theological glossary (curated renderings in every supported language) covering:
- Liturgical terms (altar, sacrament, liturgy)
- Theological concepts (salvation, grace, redemption, trinity)
- Church roles (deacon, pastor, elder, bishop)
- Sacraments (baptism, eucharist, communion)
- Worship practices (hymn, psalm, prayer, blessing)
- Biblical references (gospel, epistle, covenant)
- Denominational specifics (configurable per organization)

Organizations with theological precision enabled receive enriched glossary context.

---

## Supported Languages (Full Current List)

Danish, English, Arabic, Farsi, German, French, Spanish, Portuguese, Polish, Turkish, Ukrainian, Russian, Swedish, Dutch, Italian, Chinese (Mandarin), Korean, Japanese, Romanian, Vietnamese, Bulgarian, Hebrew, Norwegian, Czech, Finnish, Hungarian, Greek, Indonesian, Slovak, Estonian, Latvian, Lithuanian, Slovenian, Thai, Hindi, Urdu, Tagalog, Tamil, Croatian, Serbian, Albanian, Bengali, Swahili, Malay, Punjabi, Afrikaans, Azerbaijani, Belarusian, Marathi, Telugu, Gujarati, Kannada, Malayalam, Macedonian, Catalan, Galician, Basque, Bosnian, Welsh, Kazakh

Right-to-Left Languages: Arabic, Farsi/Persian, Hebrew, Urdu

Quality tiers: languages marked "OC Voice Verified" on https://ocvoice.dk/en/languages have been individually tested and fine-tuned; the rest are fully supported at high quality.

---

## Comparison with Traditional Church Translation Methods

### Traditional Interpretation Equipment

| Aspect | Traditional FM/Loop | Interpreter Booths | OCvoice |
|--------|-------------------|-------------------|---------|
| **Capital Cost** | $5,000–$20,000 | $50,000+ installation | Zero upfront |
| **Language Flexibility** | 1–3 languages max | 2–4 languages (hiring) | Dozens of languages instantly |
| **Interpreter Required** | Yes, per language | Yes, per language | AI-powered (no staff) |
| **Setup Time** | 30–60 minutes | Hours (booth setup) | 2 minutes (QR code) |
| **Congregant Experience** | Receiver device (intrusive) | Radio/earpiece dropout | Personal phone (familiar) |
| **Accessibility** | Requires special receiver | Difficult with mobility | Works on any smartphone |
| **Cost Scaling** | Linear (per language) | Exponential (interpreters) | Dramatically cheaper (per-minute) |
| **Theological Quality** | Human judgment | Variable quality | AI + glossary enforced |
| **Remote Participation** | Not possible | Not possible | Full support (same experience) |

### Why OCvoice Wins

1. **Instant Multilingual Support**: No need to hire interpreters or train staff in multiple languages
2. **Mobile-First Design**: Congregants use their own phones — no special equipment checkout
3. **Dramatically Lower Costs**: Dramatically cheaper than traditional interpretation methods
4. **Consistent Quality**: AI translation with theological glossary ensures consistent message across languages
5. **Hybrid Events**: Same experience for in-person and remote listeners (YouTube/Zoom)
6. **Professional Audio**: Premium synthesized TTS provides cleaner, more consistent audio than FM systems
7. **Accessibility**: Works with screen readers and hearing aids seamlessly

---

## Translation Quality Assurance

### Pre-Translation Safeguards

- **Profanity Filter**: Applied once upstream in transcript processing before fan-out to all languages (performance optimization)
- **Speaker Diarization**: Automatically identifies pastor vs. guest speakers, improving context
- **Code-Switching Awareness**: Native support for Danish-English mixed speech prevents ASR errors at language boundaries

### Post-Translation Monitoring

- **Session Quality Scoring**: Automated metrics post-session:
  - Translation coherence (internal consistency)
  - Glossary term hit rate (theological terms used correctly)
  - Audio quality (no dropouts, natural pacing)
  - Listener engagement (implicit quality signal)

- **Listener-Aware Translation Gating**: Skip expensive translation/TTS for languages with zero active listeners (performance + cost optimization)

### Human Oversight

Organizations can:
- View live transcript in dashboard during service
- Review quality scores after session
- Report translation errors for model improvement
- Export all transcripts and translations for archival

---

## Privacy & Data Handling

### Data Retention

- **Transcripts**: Stored indefinitely (for replay, re-translation, licensing)
- **Translations**: Stored indefinitely (for listener access, archival)
- **Audio Files**: Stored in file storage with optional cleanup
- **Listener IDs**: Hashed, no personal data collected
- **No Training Data**: OCvoice audio is NOT used to train any public AI models

### Encryption & Security

- **HTTPS Only**: All communication encrypted in transit
- **SOC 2 Type II Compliant**: Backend audited for security controls
- **No Third-party Sharing**: Audio stays within OCvoice infrastructure — no data brokers
- **GDPR Compliant**: Data deletion on demand, consent-based processing
- **Enterprise Auth**: SAML/SSO for institutional control

### Compliance

- **SOC 2 Type II**: Backend audited for security controls
- **GDPR**: European data residency options available
- **HIPAA**: Available for healthcare facilities (contact sales)
- **COPPA**: Compliance for youth ministry events

---

## Organization Settings & Configuration

### Translation Quality Settings

Organizations can customize translation behavior for their specific needs and budget:
- Automatic routing (default) — Balances quality and cost across languages
- Cost-optimized mode — Faster, more economical option
- Quality-optimized mode — Highest quality translations
- Theological precision mode — Maximum theological accuracy with enriched glossary

### Voice & Audio Configuration

- **Voice Gender**: Female or male voice per organization
- **Quality Tiers**: Organizations can choose audio quality levels
- **Per-tier Voice Preferences**: Save voice choices, restore on switching

### Worship Mode (Music Detection)

Organizations can configure auto-detection of worship music vs. sermon:
- **Off**: Translate everything (default)
- **Pass through**: Detect music, pass source text through (cost savings)
- **Skip**: Detect music, skip translation entirely
- **Adjustable Sensitivity**: Customize thresholds for high-volume audio detection and speech vs. music detection

### Source Language Configuration

- **Primary Language**: Danish (highest ASR bias) — required for streaming
- **Code-switching Support**: English and other languages mixed in sermon (e.g., `sourceLanguages: ["da", "en"]`)
- Speech recognition uses primary language as main hypothesis, secondary languages as code-switching hints

---

## Billing & Cost Structure

### Simple Pay-as-you-go Pricing

- Charged only during active sessions
- Soft-billing: charge continues during short pauses (worship mode respects pause timing)
- Volume discounts apply automatically at higher usage tiers
- Visible cost breakdown per service component and language

### Cost Components

Translation costs depend on:
- Minutes of audio and speaker count
- Words in transcript and target languages
- Characters × languages × voice quality tier
- Storage of transcripts and audio retention

### Example Value Proposition

Traditional professional interpreters for a 30-minute service with 3 languages can cost $300+. OCvoice dramatically reduces this cost through AI automation while maintaining theological accuracy.

### Free Trial

- 60 minutes of free service per organization
- After trial, pay-as-you-go (no contracts, cancel anytime)

### Pricing Information

Visit https://ocvoice.dk/en/pricing for current rates and detailed pricing structure.

---

## Setup & Integration Guide

### Quick Start (5 minutes)

1. **Create Organization**: Sign up → auto-create organization account
2. **Configure Languages**: Select any number of target languages for translations
3. **Generate QR Code**: Platform auto-generates broadcaster + listener QR codes
4. **Test Stream**: Start test session, verify audio + translation quality
5. **Go Live**: Share listener QR code → congregants scan and pick their language

### Integration Scenarios

**On-Premise**: PWA (Progressive Web App) works offline-first, no installation required
**YouTube Live**: Embed subtitle overlay or audio track (export translations → use as captions)
**Zoom/Teams**: Share screen with listener page → participants see/hear translations
**Physical Signage**: Display listener QR on projector or printed cards during service

### Broadcaster Requirements

- **Microphone**: Any USB/headset microphone (webcam built-in OK)
- **Internet**: 2+ Mbps upload for real-time streaming
- **Browser**: Chrome, Firefox, Safari (modern versions)
- **Permission**: Browser microphone + speaker access (standard PWA prompts)

### Listener Requirements

- **Smartphone**: Any iOS/Android with web browser
- **Internet**: 1+ Mbps (cellular data or WiFi)
- **Browser**: Any modern web browser (no app install)
- **Optional**: AirPods or wired earbuds for discreet listening

### Troubleshooting

| Issue | Solution |
|-------|----------|
| No audio from broadcaster | Check microphone permissions in browser, test audio input device |
| Translations delayed | Verify internet speed (2+ Mbps) and language selection |
| Listeners can't join | Verify QR code is active session, check network connectivity |
| Wrong translation language | Listener can switch language mid-stream (dropdown on listener page) |
| TTS audio sounds robotic | Switch to higher audio quality option or enable voice enhancement |

---

## Use Cases & Success Stories

### Multilingual Congregations

**Problem**: 40% of congregants speak Mandarin Chinese or Korean — need translated sermon without hiring interpreters
**Solution**: Activate Chinese + Korean target languages, share listener QR code
**Outcome**: Attendance +25%, congregant satisfaction +40%, zero interpretation overhead

### International Missionary Event

**Problem**: Visiting pastor from Ghana + Swahili-speaking congregants + European visitors — need real-time translation to 5 languages
**Solution**: Configure Danish (source) → English, Swahili, German, French, Spanish (targets)
**Outcome**: Unified experience, no equipment rental or interpreter coordination

### Hybrid Church Service

**Problem**: In-person congregation + YouTube livestream need different audio/subtitles
**Solution**: Stream to YouTube with embedded subtitles, in-person listeners use PWA
**Outcome**: Same theological message, both groups fully engaged

### Immigration Settlement Services

**Problem**: Refugee resettlement organization hosts weekly orientation in English but attendees speak 8 different languages
**Solution**: Deploy OCvoice for 8-language simultaneous interpretation
**Outcome**: Dramatically improved comprehension, faster integration, lower staff overhead

### Biblical Scholarship Seminar

**Problem**: International scholars from Denmark, Germany, Italy, Spain need simultaneous translation of Danish biblical lectures
**Solution**: Configure 4-language translation with quality settings optimized for theological precision
**Outcome**: Precise theological language, reduced post-seminar clarification needed

---

## Frequently Asked Questions

### Technical

**Q: How fast is the translation latency?**
A: End-to-end latency is typically 8–12 seconds from speaking to translation delivery. Listeners see subtitles first, audio follows.

**Q: Does OCvoice work offline?**
A: No, real-time translation requires internet connectivity. However, the listener PWA is offline-first — if connection drops temporarily, cached translations display.

**Q: Can I record the translated service for archive?**
A: Yes — all transcripts and translations are stored indefinitely. You can export full transcripts in all languages or download audio files for each language.

**Q: Does OCvoice train on our data?**
A: No. Your audio is never used to train any public AI models. It's used only for real-time translation and quality monitoring within OCvoice.

**Q: What happens if the internet cuts out mid-service?**
A: The broadcaster stops transcribing (real-time only). Listeners see a "connection lost" message. Service resumes when connection restores — no data loss.

### Commercial

**Q: What's the minimum contract?**
A: No minimum contract. Pure pay-as-you-go, cancel anytime. Free 60-minute trial to test.

**Q: Do you offer bulk discounts?**
A: Yes — volume discounts apply automatically at higher monthly usage. Contact sales for enterprise bulk rates.

**Q: Can we use our own translation/audio engines?**
A: Custom integrations available for enterprise customers (contact sales).

**Q: Is there a mobile app?**
A: No app required — Progressive Web App (PWA) works in any browser and can be added to home screen.

### Theological/Liturgical

**Q: How does OCvoice handle liturgical terms?**
A: A fully stocked theological glossary ensures terms like "sacrament," "liturgy," "eucharist," "doxology," etc. translate correctly per your tradition.

**Q: Can we add our own glossary terms?**
A: Yes — organizations can manage custom glossary entries in the dashboard (admin only).

**Q: Does OCvoice respect different Christian traditions?**
A: Yes — glossary is customizable per organization. Catholic, Orthodox, Lutheran, evangelical terms can all be configured separately.

**Q: Can OCvoice handle Scripture in original languages (Greek/Hebrew)?**
A: Yes — if pastor reads bilingual Scripture, the code-switching model preserves transliteration. Translation is context-aware for biblical terms.

---

## Roadmap & Future Features

### Coming Soon

- **Voice Cloning**: Organizations can upload pastor's voice for more natural audio synthesis
- **Custom Models**: Fine-tuned translation models for denominational-specific theology
- **Hearing Loop Integration**: Direct audio output to venue hearing loop systems
- **Mobile App**: Native iOS/Android app for offline-first listening
- **Multilingual Broadcaster**: Pastor speaks in multiple languages, OCvoice manages language boundaries
- **Collaborative Glossaries**: Denominations can share glossaries (Lutheran, Catholic, Orthodox, etc.)

### Research

- **Real-time Quality Scoring**: In-session translation quality feedback for broadcasters
- **Speaker Emotion Detection**: Preserve tone in translation (serious vs. celebratory sermon)
- **Liturgical-Aware Translation**: Context model trained on hymns, creeds, prayers
- **Low-Bandwidth Mode**: Adaptive quality for poor internet conditions

---

## Contact & Support

- **Website**: https://ocvoice.dk
- **For Churches**: https://ocvoice.dk/en/for-churches
- **Pricing & Billing**: https://ocvoice.dk/en/pricing
- **Language Support**: https://ocvoice.dk/en/languages
- **FAQ**: https://ocvoice.dk/en/faq
- **About**: https://ocvoice.dk/en/about
- **Email Support**: Available through dashboard (authenticated users)
- **Status Page**: Real-time uptime monitoring and incident reporting

---

## Company Information

**OCplan ApS**
- CVR: 42665797
- Headquarters: Denmark
- Founded: 2024
- Mission: Enable global church fellowship through real-time AI translation
- Compliance: SOC 2 Type II, GDPR, HIPAA-ready

---

**Last Updated**: April 2026
**Version**: 1.0 (OCvoice LLM Discoverability Guide)