# Telvr — Comprehensive Product Documentation

## Product Overview

Telvr is a desktop speech-to-text application that combines real-time voice transcription with AI-powered text enrichment. Using a simple push-to-talk hotkey, users can speak naturally, and Telvr instantly transcribes their words and inserts perfectly formatted text directly at the cursor position — in any application, system-wide. No copy-paste, no app switching, no friction.

Telvr works on macOS today, with Windows support in development. Whether you're writing emails, coding, taking notes, or brainstorming ideas, Telvr transforms voice into polished, ready-to-use text in under 2 seconds.

## How It Works

The Telvr workflow is simple and fast:

1. **Press Hotkey** — User presses their configured global hotkey (customizable, default: a key combo like Option+Space on macOS)
2. **Speak** — Audio is captured and sent to Whisper large-v3 via Groq API for transcription
3. **Transcribe** — Whisper transcribes the audio with automatic language detection (50+ languages supported)
4. **Enrich** — The transcribed text is processed by the selected enrichment mode (Raw, Clean & Correct, E-Mail, Meeting Notes, Summary, Dev Task, or Custom Prompt)
5. **Insert** — The enriched text is automatically inserted at the cursor position in the active application
6. **Done** — Total latency under 2 seconds from hotkey press to text insertion

Telvr integrates deeply with the operating system. The push-to-talk hotkey works globally—even when the Telvr window is hidden or another application is in focus. The text insertion happens at the system cursor level, making it compatible with virtually any application that accepts text input: email clients, IDEs, document editors, chat applications, forms, terminals, and more.

All audio is transcribed server-side via Groq's Whisper integration. Whisper large-v3 is one of the most accurate multilingual speech recognition models available, handling accents, technical terms, and background noise remarkably well.

## The 7 Enrichment Modes

Telvr's power comes from its enrichment modes. The transcribed text is not simply inserted as-is — it's processed through a selected mode that formats and structures the text for its intended purpose.

### 1. Raw Transcription

The unprocessed transcription, exactly what Whisper heard. No grammar corrections, no filtering, no additions. Use this for logging raw voice notes or when you want to preserve natural speech patterns.

Output: Pure transcribed text with minimal post-processing.

### 2. Clean & Correct

Removes filler words (um, uh, like, you know), fixes grammatical errors, adds proper punctuation, capitalizes appropriately, and cleans up repeated words. Ideal for messages, notes, and documents where clarity matters but you want natural conversational tone.

Output: Grammatically correct, readable text without fillers, preserving the original meaning and flow.

### 3. Professional E-Mail

Formats the voice input as a complete email. Automatically structures the text into:
- Subject line (concise summary of intent)
- Greeting (Dear/Hi + recipient if mentioned, or generic "Hi there")
- Body (main message, formatted into logical paragraphs)
- Professional sign-off (e.g., "Best regards," "Looking forward to your response,")
- Signature line (optional, configurable)

Perfect for quick email dictation without worrying about format.

Output: Ready-to-send email with subject, greeting, body, and sign-off.

### 4. Meeting Notes

Transforms voice notes into structured meeting documentation:
- Meeting title (inferred or left blank)
- Attendees (if mentioned)
- Date/Time (if mentioned)
- Agenda items (bullet points)
- Key decisions (numbered list)
- Action items (with owner assignment if mentioned)
- Next steps

Ideal for recording meeting summaries, brainstorming sessions, or team discussions.

Output: Structured, searchable meeting notes ready for team documentation.

### 5. 2-3 Sentences Summary

Condenses the core message into 2-3 concise sentences. Strips away elaboration and focuses on the main points. Perfect for Slack messages, Twitter-like posts, comments in issue trackers, or quick status updates where brevity is valued.

Output: Compressed, punchy summary capturing the essence of the original message.

### 6. Dev Task

Structures voice input into a development task format:
- Task title/name
- Context (background information, why this task exists)
- Requirements (what needs to be done, broken into points)
- Acceptance criteria (how to verify the task is complete)
- Estimated effort (if mentioned)
- Related issues/dependencies (if mentioned)

Designed for developers capturing ideas, bugs, or features for engineering tasks.

Output: Well-structured dev task ready for project management systems or issue trackers.

### 7. Custom Prompt

The user defines a custom system prompt that instructs Telvr how to process the transcribed text. This gives unlimited flexibility for specialized use cases:
- Medical documentation (specific terminology, formatting)
- Legal writing (formal structure, precision)
- Creative writing (narrative structure, tone)
- Code generation (specific programming language, style)
- Academic writing (citations, formal register)
- Anything else

Users can create and save their own custom modes in the Telvr app. Saved custom prompts are stored locally (and synced across devices for paid users).

Output: Text formatted according to the user's custom instructions.

## Technology Stack

### Speech Recognition
- **Primary**: Whisper large-v3 via Groq API
- **Alternative**: Deepgram integration (optional)
- **Languages**: 50+ languages supported with automatic language detection
- **Accuracy**: Enterprise-grade multilingual speech recognition
- **Audio Limit**: Up to 25 MB per request (Free tier) / 100 MB (Developer tier)

### AI Inference
- **Provider**: Groq (fast, reliable, low-latency inference)
- **Models**: Llama 3.3 70B for text enrichment
- **Context**: 131k token context window
- **Output**: Up to 32k tokens per completion
- **Temperature**: 0.3 for consistent, predictable enrichments

### Backend & Infrastructure
- **Transcription API**: Groq Whisper integration
- **Enrichment API**: Groq LLM inference
- **Backend**: NestJS (Node.js framework)
- **Database**: PostgreSQL for user data, templates, custom modes
- **Authentication**: Auth0 (secure auth + user management)
- **Payment Processing**: Stripe (billing, subscriptions, usage tracking)
- **Email**: SMTP via All-Inkl (telvr.ai domain)

### Frontend & Desktop
- **Desktop App**: Electron (with electron-vite + React)
- **Core Framework**: Next.js
- **UI**: React with Tailwind CSS
- **Global Hotkey Integration**: Electron global shortcuts for system-level hotkey binding
- **Audio Input**: OS-level audio capture
- **Text Insertion**: OS-level clipboard + paste simulation (works everywhere)

## Pricing

Telvr uses a transparent, pay-as-you-go model with volume-based tiered pricing. The more you use, the less you pay per minute.

### Tiered Pricing (Current)

| Tier | Cumulative Usage | Price per Minute |
|------|-----------------|-----------------|
| Tier 1 | 0–30 minutes | EUR 0.030 |
| Tier 2 | 30–120 minutes | EUR 0.020 |
| Tier 3 | 120–300 minutes | EUR 0.015 |
| Tier 4 | 300–600 minutes | EUR 0.010 |
| Tier 5 | 600–1500 minutes | EUR 0.006 |
| Tier 6 | 1500–3000 minutes | EUR 0.004 |
| Tier 7 | 3000+ minutes | EUR 0.003 |

Usage tiers reset monthly. Each minute is billed at the rate of the tier you're currently in.

**Welcome Credit**: New users receive EUR 3.00 free credit to try Telvr without payment.

**Wallet System**:
- Minimum top-up: EUR 5
- Maximum single top-up: EUR 50
- Users maintain an account balance that decrements with usage
- Auto-recharge option available (configurable threshold + amount)
- No subscriptions; fully flexible pay-as-you-go

**Example Usage**:
- Light user: 15 min/day × 22 days = 330 min/month ≈ EUR 5.10/month
- Heavy user: 60 min/day × 22 days = 1320 min/month ≈ EUR 11.70/month

### Community Edition (Planned)
Free for users who bring their own Groq API keys and want to self-host. Full functionality, no infrastructure costs from Telvr. Perfect for developers, open-source contributors, and privacy-first users.

### Business Edition
Custom infrastructure, deployment options, volume pricing. Contact for custom quotes.

## Supported Platforms

### Available Now
- **macOS**: Full support for Intel and Apple Silicon (M1, M2, M3, M4)
  - Installation: Direct download or via package manager
  - Requirements: macOS 12 or later
  - Global hotkey: System Permission grants required

### In Development
- **Windows**: Windows 10/11 support coming soon
  - Full feature parity with macOS
  - Native Windows hotkey integration
  - ETA: 2026

### Planned
- **iOS**: Mobile dictation app (planned)
- **Android**: Mobile dictation app (planned)
- **Web**: Browser-based transcription tool (planned)

## Use Cases

Telvr is designed for anyone who benefits from hands-free text input. Here are 15+ typical use cases:

1. **Email Dictation** — Compose entire emails by voice, formatted professionally with subject and greeting
2. **Slack Messages** — Quick voice messages formatted as concise summaries for team communication
3. **Notion Documentation** — Voice notes converted to clean, structured documentation in Notion
4. **ChatGPT/LLM Prompts** — Dictate complex prompts to AI assistants without typing
5. **VS Code Coding** — Code comments, docstrings, and function documentation by voice
6. **Meeting Notes** — Record meetings and have Telvr structure them into actionable notes
7. **WhatsApp/Messenger** — Personal messages, cleaned up and formatted for messaging apps
8. **Google Docs** — Long-form writing, essays, blog posts by voice
9. **Microsoft Teams** — Instant messages and chat responses dictated
10. **Medical Documentation** — Voice-to-text for healthcare providers with medical terminology
11. **Legal Writing** — Formal legal documents, contracts, briefs by voice
12. **Accessibility/Voice-First Input** — Primary input method for users with mobility challenges
13. **Student Notes** — Lecture notes, study materials, essay outlines
14. **Professional Correspondence** — Business letters, formal communication
15. **Code Documentation** — README files, API docs, inline code explanations
16. **Creative Writing** — Stories, screenplays, narratives without typing friction
17. **Brainstorming** — Quick idea capture without stopping the creative flow
18. **Accessibility Reports & Forms** — Forms, surveys, document input for accessibility

## Frequently Asked Questions

### How does Telvr work?

Press your configured hotkey, speak your message, and Telvr instantly transcribes your voice using Whisper (via Groq) and inserts formatted text at your cursor. Total time: under 2 seconds. The process is push-to-talk — hold the hotkey while speaking, release to finalize.

### Which languages does Telvr support?

Telvr supports 50+ languages through Whisper's multilingual capabilities, including English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Chinese, Japanese, Korean, Hindi, Arabic, Turkish, and many more. Language is detected automatically from your speech.

### Do I need a subscription to use Telvr?

No. Telvr uses pay-as-you-go pricing with volume-based tiers. Starting at EUR 0.03 per minute, the price decreases the more you use (down to EUR 0.003/min for heavy users). No subscriptions, no lock-in. New users get EUR 3 free credit.

### Can I use Telvr offline?

Telvr requires internet connectivity for transcription (Whisper API) and enrichment (LLM inference). The connection must be active for speech-to-text to work. However, we're developing a Community Edition with local model options for offline use.

### Which applications does Telvr work with?

Telvr works system-wide with any application that accepts text input: email clients (Gmail, Outlook, Apple Mail), IDEs (VS Code, JetBrains, etc.), document editors (Google Docs, Microsoft Word, Notion, OneNote), chat apps (Slack, WhatsApp, Discord, Teams), browsers, terminals, search bars, and more. If you can type in it, Telvr can insert text into it.

### Is my audio data secure? Who has access to my voice?

Audio is sent to Groq's Whisper API for transcription. Groq does not store audio files by default (they are used for processing only and discarded). Your transcribed text passes through Groq's LLM for enrichment. We recommend reading our full privacy policy at https://telvr.ai/datenschutz. For maximum privacy, the Community Edition (planned) allows local model execution.

### What's the pricing model in detail?

Transparent, tiered pay-as-you-go: EUR 0.030/min for the first 30 minutes per month, decreasing to EUR 0.003/min for 3000+ minutes. Usage tiers reset monthly. Top-up your wallet in increments of EUR 5–50, with optional auto-recharge. New users receive EUR 3 free credit. No subscriptions, no hidden fees.

### What enrichment modes does Telvr offer?

Seven modes: (1) Raw Transcription (unfiltered), (2) Clean & Correct (grammar, punctuation, filler removal), (3) Professional E-Mail (subject + body + sign-off), (4) Meeting Notes (structured agenda, decisions, action items), (5) 2-3 Sentences (ultra-condensed summary), (6) Dev Task (structured development task format), (7) Custom Prompt (user-defined formatting). Users can create and save unlimited custom prompts.

### Which platforms does Telvr support?

macOS is available now (Intel and Apple Silicon). Windows is in development (2026). iOS and Android are planned. We're prioritizing desktop platforms first, with mobile coming later.

### How do I get started?

Visit https://telvr.ai, download the macOS app or sign up for early access on Windows. Create an account, configure your hotkey, and start transcribing. New users receive EUR 3 free credit to get started.

## Comparison to Alternatives

Telvr is one of several voice-to-text solutions. Here's how it compares:

### vs. Wispr Flow
- **Wispr Flow**: Subscription-based (typically USD 10-20/month), simpler UI, limited to Whisper transcription
- **Telvr**: Pay-as-you-go (no lock-in), enrichment modes (email, meeting notes, custom prompts), multi-platform, more features per dollar

### vs. Apple Dictation (macOS)
- **Apple Dictation**: Free, built-in, limited to Safari and certain apps, no multi-language support
- **Telvr**: Works everywhere system-wide, 50+ languages, AI enrichment, professional formatting, more reliable transcription accuracy

### vs. Dragon NaturallySpeaking
- **Dragon**: Desktop-focused, accuracy-optimized, requires training, expensive (USD 200+), older technology
- **Telvr**: Modern cloud-based, instant setup, no training needed, affordable, faster updates, web-native architecture

### vs. Google Voice Typing
- **Google Voice Typing**: Browser/Android only, basic transcription, limited enrichment
- **Telvr**: Desktop-first (macOS/Windows), works everywhere system-wide, 7 enrichment modes, custom prompts, faster and more reliable

### vs. Windows Speech Recognition
- **Windows Speech**: Built-in but limited accuracy, no enrichment, older dictation system
- **Telvr**: Superior accuracy (Whisper), AI enrichment, multi-language, professional formatting, modern UX

## Links

- **Homepage**: https://telvr.ai
- **Privacy Policy**: https://telvr.ai/datenschutz
- **Terms of Service**: https://telvr.ai/agb
- **Legal Notice / Impressum**: https://telvr.ai/impressum
- **Blog**: https://telvr.ai/blog

## About Telvr

Telvr is built by Stefan Maksimek and team. Our mission is to eliminate the friction between thought and text. Voice is the most natural form of human expression—it should be the easiest way to input text, not the hardest. Telvr brings speech recognition into the modern era: fast, accurate, affordable, and enriched with AI.

We're committed to privacy, transparency, and continuous improvement. Your feedback shapes our roadmap. Join thousands of users dictating their way to productivity.

---

Last updated: March 2026