walkie-talkie

What This Skill Does

The walkie-talkie skill transforms your WhatsApp experience by enabling a seamless voice-to-voice communication loop. Instead of relying on traditional text messaging, this skill allows the OpenClaw agent to process incoming audio messages, transcribe them into text, generate an intelligent response, and convert that response back into a natural-sounding voice note. Designed for high performance, it utilizes local processing tools like whisper-cpp and sherpa-onnx-tts to ensure that conversations feel instantaneous, maintaining a Real-Time Factor (RTF) of less than 0.5 seconds. By providing both the text transcript and the audio file, the skill ensures that clarity is never sacrificed for convenience, catering to users who prefer hands-free interaction or simply want a more conversational feel.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/rubenfb23/walkie-talkie-mode

Ensure that you have the necessary system dependencies installed, specifically ffmpeg, whisper-cpp, and sherpa-onnx-tts, as these are critical for the skill's local processing requirements. Once installed, the skill will automatically monitor incoming WhatsApp messages for audio files or specific trigger phrases.

Use Cases

This skill is perfect for situations where typing is inconvenient or impossible, such as while driving, walking, or engaging in multi-tasking. It is also ideal for users with accessibility needs who prefer voice interfaces. Whether you are conducting a quick status update meeting over WhatsApp or simply catching up with friends, the walkie-talkie mode makes your agent a versatile, interactive voice assistant.

Example Prompts

1. "Activa modo walkie-talkie, necesito que me escuches mientras conduzco."
2. "Hablemos por voz de ahora en adelante, responde a mis notas de audio."
3. "Activa el modo walkie-talkie ahora."

Tips & Limitations

To maintain optimal performance, ensure your device has sufficient CPU overhead to handle the transcription and synthesis tasks concurrently. Since this skill relies on local processing, response quality is hardware-dependent. Always verify that your microphone permissions are granted to the OpenClaw agent to avoid errors. Note that the skill works best in quiet environments; loud background noise may impact the accuracy of the transcription process. If the agent struggles to transcribe your audio, consider speaking clearly and slightly slower than in standard conversation.