Analysis updated 2026-05-18
Get a spoken alert when a background coding agent finishes its task
Hear permission requests read aloud instead of missing them on screen
Stay informed about errors or rate limits without watching the terminal
Use a fully local, no API key voice notification tool with a coding agent
| pg-parunson/crier | 0-bingwu-0/live-interpreter | 0cm-labs/tokenizer-benchmark | |
|---|---|---|---|
| Stars | 2 | 2 | 2 |
| Language | Python | Python | Python |
| Setup difficulty | easy | moderate | moderate |
| Complexity | 2/5 | 2/5 | 2/5 |
| Audience | developer | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
Currently macOS only, and requires restarting your coding agent session after install.
Crier is a small tool that gives a spoken voice to AI coding assistants like Claude Code, Codex, Gemini, or Cursor. Instead of you having to keep checking back on a terminal to see if your coding agent has finished a task, crier speaks a single sentence out loud when something actually needs your attention, so you can stay focused on other work and just listen for it. It only speaks at four moments: when a task finishes, when the agent needs your permission to do something, when something goes wrong, or when it has been sitting idle waiting for you. Everything else stays silent, including routine background activity, so it does not become an annoying stream of notifications. It never reads out the actual answer, since you can read faster than you can listen, and if you start typing it stops talking immediately. The sentence you hear is written by the coding agent itself as one extra line at the end of its response, so there is no second AI model or separate API call involved, and no additional cost. For permission requests, it reuses the plain language description the agent already writes next to a command, so instead of a vague prompt you hear something specific like being asked whether it is fine to delete a folder. A short chime plays just before each spoken line so a voice does not startle you out of nowhere. Everything runs locally on your own machine using an open source speech model, so there is no API key, no account, and no per use billing. It supports Korean, English, and Japanese, with several voice options and adjustable tone. Installation is a single command line script, currently for macOS, and a built in diagnostic command helps figure out what is wrong if audio does not play. This tool suits developers who regularly leave coding agents running in the background and want to know, without staring at a screen, exactly when their attention is actually needed. It is released under the MIT license.
A local tool that gives coding AI agents a spoken voice, announcing in one sentence only when a task finishes, needs permission, or fails.
Mainly Python. The stack also includes Python, ONNX.
You can use, modify, and distribute this freely, including in commercial projects, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.