Varterm — Free Text to Speech Reader

Convert any text to natural-sounding speech with premium AI voices. Free to use, with optional upgrades for HD quality.

Features

• 🆓 Free tier with browser-native voices
• 🎙️ Premium voices via ElevenLabs integration
• 💳 Stripe billing for subscriptions
• 🤖 ChatGPT integration via Custom GPT Actions
• 🔧 Claude integration via MCP (Model Context Protocol)
• ⚡ Vercel-ready for instant deployment

Quick Start

# Clone the repo
git clone https://github.com/varterm/platform.git
cd varterm

# Install dependencies
npm install

# Copy environment variables
cp .env.example .env.local

# Run development server
npm run dev

Open http://localhost:3000 to see the app.

Open Source Project Setup

• LICENSE uses MIT for permissive reuse.
• CONTRIBUTING.md explains setup, checks, and PR expectations.
• CODE_OF_CONDUCT.md sets collaboration standards.
• SECURITY.md explains private vulnerability reporting.
• .github/ISSUE_TEMPLATE includes structured bug and feature forms.
• .github/pull_request_template.md standardizes PR quality.
• .github/workflows/ci.yml runs lint and tests on PRs/pushes.
• Extension release automation lives in the varterm/extensions repository.

Binary build artifacts are intentionally not committed to git.

Environment Variables

Create a .env.local file with:

# Required for premium voices
ELEVENLABS_API_KEY=your_elevenlabs_api_key

# Optional: AI answers for /api/ask
ANTHROPIC_API_KEY=your_anthropic_api_key

# Optional: protect /api/ingest and /api/ask with bearer auth
VARTERM_EXTENSION_API_TOKEN=your_private_token

# Optional: Stripe for billing
STRIPE_SECRET_KEY=sk_live_...
STRIPE_WEBHOOK_SECRET=whsec_...

# Your domain
NEXT_PUBLIC_BASE_URL=https://varterm.com

Deploy to Vercel

Or via CLI:

npm i -g vercel
vercel

Set Up Stripe Billing

1. Go to Stripe Dashboard → Payment Links
2. Create 3 recurring products:
   • Starter: $5/month
   • Pro: $15/month
   • Unlimited: $29/month
3. Copy each payment link URL
4. Update STRIPE_LINKS in app/page.js:

const STRIPE_LINKS = {
  starter: 'https://buy.stripe.com/your_starter_link',
  pro: 'https://buy.stripe.com/your_pro_link',
  unlimited: 'https://buy.stripe.com/your_unlimited_link',
};

API Endpoints

GET /api/tts

List available voices.

POST /api/tts

Convert text to speech.

Request:

{
  "text": "Hello world",
  "voice": "Sarah",
  "speed": 1.0
}

Response: audio/mpeg binary data

GET /api/mcp

MCP server info and capabilities.

POST /api/mcp

Handle MCP tool calls for Claude integration.

POST /api/ingest

Create a temporary document session for long text/script ingestion with chunking limits.

Request:

{
  "documents": [
    {
      "path": "src/app.ts",
      "content": "very long file text..."
    }
  ],
  "options": {
    "chunkSize": 1800,
    "overlap": 250,
    "maxChunks": 2000,
    "maxTotalChars": 1000000
  }
}

POST /api/ask

Ask a question against an ingest session.

Request:

{
  "sessionId": "uuid",
  "question": "Where is auth validated?",
  "maxChunks": 8,
  "maxContextChars": 15000
}

---

ChatGPT Integration (Custom GPT with Actions)

1. Go to chat.openai.com/gpts/editor
2. Create a new GPT
3. Add these instructions:

You have access to Varterm, a text-to-speech service. When users want to hear text read aloud, use the textToSpeech action. You can also list available voices with listVoices.

Always ask what voice they prefer if they don't specify one. Popular choices are Sarah (warm), Charlie (friendly), and Rachel (ultra realistic).

4. In the Actions section, click "Create new action"
5. Import the OpenAPI schema from: https://varterm.com/openapi.json
6. Save and publish your GPT

Example GPT Conversation

User: Read this paragraph aloud using a British voice

GPT: I'll use George's British voice for this. [calls textToSpeech]

Here's the audio: [plays audio]

---

Claude Integration (MCP Server)

Option 1: HTTP-based MCP (Recommended)

Add this to your Claude Desktop config (~/.config/claude/claude_desktop_config.json):

{
  "mcpServers": {
    "varterm": {
      "transport": {
        "type": "http",
        "url": "https://varterm.com/api/mcp"
      }
    }
  }
}

Option 2: Local MCP Server

For local development or if you have your own ElevenLabs key:

1. Install MCP SDK:

npm install @modelcontextprotocol/sdk

2. Add to Claude Desktop config:

{
  "mcpServers": {
    "varterm": {
      "command": "node",
      "args": ["/path/to/varterm/mcp-server.js"],
      "env": {
        "ELEVENLABS_API_KEY": "your_key_here"
      }
    }
  }
}

3. Restart Claude Desktop

Available MCP Tools

Tool	Description
text_to_speech	Convert text to audio with optional voice and speed
list_voices	List all available voices

Example Claude Conversation

User: Can you read this article summary aloud?

Claude: I'll convert that to speech for you. [uses text_to_speech tool]

Done! I've generated audio using Sarah's voice. The audio file is ready.

---

Project Structure

varterm/
├── app/
│   ├── api/
│   │   ├── tts/
│   │   │   └── route.js      # TTS API endpoint
│   │   └── mcp/
│   │       └── route.js      # MCP endpoint for Claude
│   ├── globals.css
│   ├── layout.js
│   ├── page.js               # Main app
│   └── page.module.css
├── lib/
│   └── elevenlabs.js         # ElevenLabs integration
├── public/
│   └── openapi.json          # OpenAPI spec for GPT Actions
├── mcp-server.js             # Standalone MCP server
├── package.json
└── README.md

Pricing Tiers

Plan	Price	Characters	Voices
Free	$0	Unlimited	Browser voices
Starter	$5/mo	50,000	10 premium
Pro	$15/mo	200,000	30+ premium
Unlimited	$29/mo	Unlimited	All 50+ Ultra HD

Tech Stack

• Framework: Next.js 14
• TTS Provider: ElevenLabs
• Payments: Stripe Payment Links
• Deployment: Vercel
• AI Integrations: OpenAI GPT Actions, Claude MCP

License

MIT © Varterm

---

Built with ❤️ for readers everywhere.