FOSSWire
Search/
Open Hardware & RISC-V

Walkie Claudie: Open-source voice interface for Claude Code on mobile

Brief written by the FOSSWire AI editor from Hackaday How we write

1 min read

Key Takeaway

A delivery driver built Walkie Claudie, an open-source push-to-talk voice interface that lets users interact with Claude Code via speech recognition and text-to-speech on a smartphone connected to a PC.

Summary

Developer [Bookmari] created Walkie Claudie to solve a practical problem: using Claude Code while working as a delivery driver, where typing prompts and reading responses was impractical. The system runs on a smartphone and connects directly over the network to a PC running Claude Code, with VPN support for remote use. It features speech recognition for input and text-to-speech for Claude's responses. A push-to-talk mechanism is triggered by a button on Bluetooth earbuds, and a special mode maintains the connection during long pauses for extended thinking. The interface is styled with retro walkie-talkie aesthetics, including synthesized squelch and channel-switching sound effects.

How Walkie Claudie works

  • Architecture: The smartphone client connects directly over the network to a PC running Claude Code, with VPN as an option for non-local connections rather than routing through cloud APIs.
  • Voice pipeline: Speech recognition handles user input conversion to text prompts, while a text-to-speech engine renders Claude's text responses as spoken audio.
  • Hardware integration: Push-to-talk activation uses a physical button on Bluetooth earbuds, enabling hands-free operation in mobile work environments.
  • Session management: A dedicated mode keeps the communication channel open during extended pauses, accommodating longer thinking periods without timeout disconnection.
  • Interface design: The UI mimics vintage walkie-talkie LCD displays with synthesized audio feedback including squelch tails and channel-switching tones for tactile familiarity.

Why it matters

Walkie Claudie demonstrates how open-source tooling can bridge the gap between desktop AI assistants and mobile, hands-free use cases without proprietary middleware. For developers and makers, it provides a hackable reference architecture for voice-enabling local LLM workflows, potentially inspiring similar interfaces for other coding assistants or self-hosted models. The direct PC-to-phone approach also preserves privacy by avoiding cloud-based voice API dependencies.

Original reference: hackaday.com →