Viberadio Fm

VibeRadio

Website · Setup · Library and MCP

VibeRadio is an open-source text-to-speech CLI and library that turns supplied text or coding-session events into local spoken narration for developers.

Install

npm install -g viberadio-fm

viberadio say "Build passed"

Requires Node.js 18 or newer and a system speech tool: macOS say, Linux espeak, or Linux spd-say. Success: you hear “Build passed” and see a local-provider status message. If speech is unavailable, the CLI prints the text and a setup warning, then exits successfully. No API key or environment variables are required. The CLI does not load .env files.

Recap a session

printf '[{"kind":"tests-pass","count":12}]' | viberadio recap

printf '[{"kind":"tests-pass","count":12}]' | viberadio recap --style podcast

The second command changes the script to host-labelled dialogue. Both styles use one system voice. You can also pass a JSON-array file: viberadio recap session.json --mode podcast.

Features and limits

• Local speech: the default audio path disables network egress and uses the system speech runner from @pooriaarab/vibe-core.
• Recap scripts: formats milestone events; accepts monologue or podcast styles and summary or podcast modes.
• Library bridge: successful library recap() appends a best-effort event for vibenotifications; CLI and MCP recaps call narration directly.
• Two MCP tools: narrate speaks text; recap builds and speaks an event recap.

text / event array -> narration script -> system speech tool -> audio

There is no automatic agent-session watcher or background start command. Reactive music, cloud voices, distinct podcast voices, recording, visualizers, and a menubar application do not ship. This package does not include Claude Code, Codex, or Gemini adapters; pass text or events from your workflow. Source inspection confirms three system-tool choices through the shared runner. No physical audio output was verified for this rewrite.

Library and MCP

import { buildRecapScript, narrate, recap } from 'viberadio-fm';

console.log(buildRecapScript([{ kind: 'tests-pass', count: 12 }])); await narrate('Build passed'); await recap([{ kind: 'task-done', message: 'Fixed login' }]);

buildRecapScript() is pure and works without speech tools. narrate() throws TtsUnavailableError when audio cannot run. createNarrator() exposes dependency injection; TypeScript declarations ship with the package. Configure an MCP client with:

{"mcpServers":{"viberadio":{"command":"viberadio","args":["mcp"]}}}

CLI commands: say, recap, mcp, help/--help/-h, and --version/-v. The say command accepts --style/-s; recap also accepts --mode/-m and file or stdin input. Set VIBERADIO_DEBUG=1 for speech-failure details.

Status and links

Version 0.2.4 is pre-1.0. Interfaces can change; this package provides no managed audio service. The repository's interactive prototype and launch video show product direction, not shipped functionality. Report an issue · Vibe Suite · MIT license