I built OpenVoice AI: a self-hosted AI communication platform starting with an AI phone agent
Hi everyone,
I’m building OpenVoice AI, an open-source self-hosted AI communication platform.
The project starts with an AI phone-agent stack, but the long-term goal is broader: phone calls, email, calendar, dashboards, automations and multiple AI providers in one self-hosted system.
Current v0.1.0 preview includes:
- Asterisk phone-agent baseline
- local STT/TTS foundations with faster-whisper and Piper
- FastAPI dashboard foundation
- call recording and transcript workflow
- email draft foundations
- NVIDIA / OpenRouter provider configuration
- Proxmox LXC helper-script drafts
Repo:
https://github.com/nikpottbecker/openvoice-ai
Important: this is still preview / early alpha, not a polished production release.
The biggest current challenge is real phone-call speech-to-text quality, especially German telephone audio. I’m looking for feedback from people who run Proxmox, Asterisk, SIP, FRITZ!Box, n8n, local AI or self-hosted automation stacks.
Questions:
- Would you self-host something like this?
- Which integrations would matter most to you?
- What would make the setup trustworthy enough to try?
- Which STT provider should I test next for telephone audio?
I’m the maintainer of the project. Feedback, criticism and architecture suggestions are welcome.