AI-powered voice messaging platform. Synthesize speech, clone voices, translate across languages — deliver via iOS, Telegram, or API.
From synthesis to cloning to cross-language translation — a complete voice infrastructure that runs on CPU.
Three-tier TTS: Kokoro for speed, Piper for extended languages, ElevenLabs for premium quality. All CPU-friendly.
Self-hosted Whisper with faster-whisper. Accurate transcription with automatic language detection, running on CPU with int8 quantization.
Clone any voice from a 5-second audio sample via ElevenLabs. Use cloned voices across all messaging and synthesis endpoints.
Full STT → LLM → TTS pipeline. Translate spoken audio or text across 30+ languages with automatic voice selection per target language.
Send synthesized voice messages between users. Cross-channel delivery via iOS push notifications and Telegram. Scheduled messages supported.
HMAC-SHA256 signed webhook delivery for voice events. Auto-deactivation after failures. Full CRUD management via API or MCP.
Automatic routing picks the right engine for the language and quality level. All standard engines run entirely on CPU.
82M parameter model. Apache 2.0 licensed. The default engine — fast, natural, and completely self-hosted.
ONNX-based engine. Auto-fallback for 17 additional languages not covered by Kokoro. Models downloaded lazily.
Cloud API for highest fidelity. 32 languages, voice cloning from 5-second samples, and expressive synthesis.
Native iOS app, full-featured Telegram bot, and a REST API for building custom integrations.
Native Swift app with push notification sounds. Receive voice messages as custom notification audio. Friends, inbox, and voice selection built in.
App StoreFull-featured bot: /voice, /send, /translate, /clonevoice, /ai mode, friend management. Auto-provisions users. Redis rate limiting.
Open in Telegram47+ endpoints across 14 categories. JWT, API keys, and magic codes for auth. Interactive Swagger docs. OpenAPI 3.0 spec included.
API DocumentationMagic codes for one-liner access. MCP server for AI agents. Full Swagger docs for everything else.
One-line API calls with no JWT or headers needed. Generate a MV-XXXXXX code and start calling endpoints immediately.
Model Context Protocol server exposing 7 tools — synthesize, transcribe, translate, send_voice, list_voices, clone_voice, manage_webhooks. Works with Claude Code, Cursor, and any MCP client.
Interactive API documentation with try-it-out for all 47+ endpoints. Authenticate with JWT, API key, or magic code and test directly in your browser.
Open Swagger DocsReady-made skill for the OpenClaw AI assistant. Exposes synthesize, transcribe, send_voice, list_voices, and clone_voice as AI-callable actions.
Choose the right auth level for your use case — from quick scripts to production integrations.
6-character codes for scripts and AI agents. Configurable TTL from 1 hour to 7 days. No headers required.
Long-lived keys for production integrations. SHA-256 hashed storage. Pass via X-API-Key or Authorization header.
180-day tokens from Apple or Google OAuth. Full user context, all endpoints. Used by the iOS app.