ZMachineryEquipment solutions

Technologies · voiceboi

voiceboi.

voiceboi is an on-device speech-to-speech assistant. Talk to it, and it replies as text or speech. You can clone your own voice or design a new one for it to speak with.

voiceboi is in development and not available yet.

Install: Coming soon All technologies

What it is

  • Runs on your device: Whistle speech recognition feeds the Needle tool-calling model, which feeds a native Rust port of Qwen3-TTS (Apache 2.0).
  • Replies can be text or speech.
  • Clone your own voice, or design a new one.

Voice safety

  • Voice cloning includes consent checks.
  • Voices are kept in an encrypted voice store.
  • Generated speech carries an AI-generated label and a watermark.
  • Prompts that impersonate real people are refused.

How it is built

  • A Rust workspace built on Candle.
  • A gpui-kit desktop app and a CLI.
  • An installable PWA shell.
  • Zip builds made by GitHub Actions. Downloads are not available yet (Coming soon).

Status

In development. The first stage (PR 1) is merged. The next two stages are planned and not merged yet:

  1. Planned: device and terminal control. Letting the AI control devices and terminals (PR 2).
  2. Planned: watch client. A streaming ESP32-S3 watch client over Meshtastic (PR 3).