COSS-India/VoicEra

A complete voice AI building block with telephony integration, featuring real-time speech-to-text, text-to-speech, and LLM-powered conversational agents.

What it solves

VoicEra provides an open-source infrastructure layer for building and deploying self-hosted, real-time voice AI agents, specifically targeting telephony-based interactions. It eliminates vendor lock-in by allowing operators to choose their own speech-to-text (STT), text-to-speech (TTS), large language models (LLMs), and telephony providers.

How it works

The platform separates the control plane from real-time execution. An API handles agent configuration, campaigns, and orchestration, while a dedicated Voice Runtime (powered by Pipecat) manages live call audio and model interactions. The system uses a provider-neutral architecture where AI and telephony integrations are interchangeable via stable interfaces. It can be deployed using Docker and supports both cloud APIs and locally hosted models via an optional model server.

Who it’s for

It is designed for governments, NGOs, public-interest organizations, and developers who need to deploy voice AI agents while maintaining full ownership of their data, infrastructure, and model choices, particularly for Indian and other low-resource languages.

Highlights

  • Provider Neutrality: Swap STT, TTS, LLM, and telephony providers without rewriting the platform.
  • Data Sovereignty: All call media, transcripts, and recordings are stored on the operator's own infrastructure.
  • Indic Language Support: First-class integration with Bhashini, AI4Bharat, and Kenpath Vistaar for underserved languages.
  • Self-Hosted: Fully deployable on private infrastructure to avoid reliance on managed AI services.

Related

  • Project
    BlackBeltTechnology/pi-agent-dashboard

    PI Dashboard is a web/Electron UI that lets you run, monitor, and interact with multiple *pi* AI coding‑agent sessions in real time, with features like workspace organization, token‑cost tracking, integrated terminal, OpenSpec/flow support, provider management, and extensibility via plugins.

  • Project
    huanchong-99/SoloDawn

    SoloDawn is an open‑source web app that automates full‑stack software development. It orchestrates multiple AI command‑line tools (Claude Code, Codex, etc.) via a primary LLM‑driven agent that spawns child agents on demand. Tasks are executed in parallel Git branches, each passing three layers of quality checks. When the generated code meets a 90‑point rubric, it is auto‑merged, delivering a production‑ready product from a simple natural‑language description.

  • Project
    opennookorg/betterwispr

    BetterWispr is an open‑source macOS dictation tool that transcribes speech offline using Whisper, Parakeet, or Apple’s on‑device recognizer. It lets you dictate into any app via a shortcut, supports custom vocabularies, learns from corrections, offers voice commands, and can integrate external speech‑to‑text APIs while keeping data local.

  • Project
    ManimCommunity/manim-voiceover

    A Manim plugin that integrates AI voiceovers and recording tools directly into Python, enabling per-word animation synchronization using OpenAI Whisper.