Cartesia
Low-latency speech models and voice agents
VoiceStudio
Fully local ElevenLabs alternative with an MCP server for agents
About
Open-source, local voice cloning and dubbing studio
VoiceStudio (AGPL-3.0, Electron desktop app, default engine k2-fsa/OmniVoice) runs voice work locally: clone a voice from a reference clip, design a new voice from a text description, dub videos with timed speech, dictate with a floating widget, and batch stories or audiobooks. The MCP server is mounted on the running backend at /mcp and exposes generate_speech, clone_voice, design_voice, describe_voice, transcribe, list_voices and check_health, so each agent can get its own bound voice. Audio can be returned as files on disk instead of base64 to save context. There is also a local API and an agent skill (npx skills add debpalash/VoiceStudio). Runs on Apple Silicon, NVIDIA GPUs or CPU (slower). Repo: https://github.com/debpalash/VoiceStudio.
Features
- Voice cloning from a reference recording
- Voice design from a text description
- Video dubbing with timed speech
- Transcription in 646 languages
- Built-in MCP server and local API for agents
- Agent skill via npx skills add
- Runs locally, remote services optional
Install
Related skills
All skillsRelated articles
All articlesVapi vs Retell AI (2026): Pricing, Free Tier and MCP
I compare Vapi and Retell AI for AI phone agents: per-minute pricing, free credits, concurrency, MCP servers and which one fits a solo founder.
How to List Your Product as a Grok Bot Connector
The Grok Bot marketplace now takes connector submissions. What to build, how to submit, and how bot templates turn a connector into signups.
7 Best MCP Servers for Solo Founders (2026)
Official MCP servers that let your agent read your bank, CRM, database, deployments and designs, so you can run the business by asking.
