nikpottbecker/openvoice-ai — explained in plain English
Analysis updated 2026-05-18
Run a self-hosted AI phone agent that answers calls and summarizes them.
Get AI-drafted email replies from phone call transcripts before sending them manually.
Monitor live calls, transcripts, and history through a private dashboard.
Connect a private phone system to an AI provider without relying on a cloud vendor.
| nikpottbecker/openvoice-ai | alx-code/lingbot-video-1.3b-fp8 | andrewbergman/kuma-importer | |
|---|---|---|---|
| Stars | 7 | 7 | 7 |
| Language | Python | Python | Python |
| Setup difficulty | hard | — | moderate |
| Complexity | 4/5 | — | 2/5 |
| Audience | ops devops | developer | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Needs a SIP/Asterisk phone setup, an AI provider key, and should sit behind Cloudflare Access, not exposed directly.
OpenVoice AI is a self hosted platform that ties together phone calls, email, calendars, automations, and AI providers into one system you run yourself rather than through a cloud service. The current version focuses mainly on a German language AI phone agent built on the Asterisk phone system, using local speech to text and local text to speech, which can summarize calls with AI and draft follow up emails, all viewable through a password protected web dashboard. The longer term goal described in the project is a privacy first hub that also connects to WhatsApp, Discord, Home Assistant, and workflow automation tools like n8n, though those parts are not built yet. The project is explicitly labeled an early preview: the phone agent has been tested in a real local setup, but the steps for other people to install it themselves are still being smoothed out. The dashboard, built with FastAPI and a simple web frontend, shows live call status, active calls with transcripts, call history with playback, AI response timing, drafted emails awaiting a human to send them, and settings, all without exposing sensitive keys. The project stresses that this dashboard should never be exposed directly to the internet and instead recommends Cloudflare Tunnel with Cloudflare Access, or a similar authenticated setup, in front of it. Installation targets Debian, Ubuntu, Proxmox containers, and Docker, with a setup script that copies an example environment file for you to fill in with things like your phone provider's SIP details, AI provider keys, and email credentials. It supports NVIDIA's AI service and OpenRouter as language model providers, with a modular design meant to support more providers later. Speech recognition and speech synthesis can run entirely locally, though the language model itself currently needs an external OpenAI compatible provider. The project is released under the MIT license.
A self-hosted platform that connects phone calls, email, and calendars to AI providers, currently centered on a German AI phone agent.
Mainly Python. The stack also includes Python, FastAPI, Asterisk.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.