voize

MCP server for speech generation and media URL publishing.
Generates audio from text and returns a public HTTPS URL for downstream workflows.
Built by Aliens.

Why

Voize separates speech generation + media hosting from downstream writes. Any MCP client can call one tool to generate and upload audio. The tool returns a strict JSON payload with a public URL. Callers can submit the returned URL to their own destination. This keeps credentials server-side, preserves MCP typed contracts, and avoids parsing shell output.

Getting Started · Commands · Architecture · Providers · Examples

Usage

REQUIREMENTS

What It Provides

A single STDIO MCP server built on @modelcontextprotocol/sdk with:

• generate_tts_url — the core tool for TTS generation and storage upload
• Built-in and sample-based voice-clone modes via Xiaomi MiMo
• S3-compatible object-storage upload to Cloudflare R2
• Startup guardrails for public-domain output
• Validation, retries, and structured error results

Quick Start

bun install
cp .env.example .env
bun run build
bun run build/index.js

Contributing

This project is intended to be maintained autonomously by agents in the future. Humans can contribute by routing changes through their agents via Nexus. See CONTRIBUTING.md for details.

License

MIT