videochat-withme

What This Skill Does

The videochat-withme skill transforms your OpenClaw agent into an interactive, real-time video companion. By integrating cloud-based speech recognition (Groq Whisper) and high-quality synthetic voices (edge-tts), this skill enables a fluid, natural conversation loop where your agent can "see" through your camera and "hear" through your microphone. The agent processes visual input (base64 frames) and audio transcripts through the OpenClaw chatCompletions API, allowing it to maintain memory, personality, and situational awareness during a live interaction. Whether you are at your desk or connecting remotely, this skill bridges the gap between text-based AI interactions and human-like presence.

Installation

To get started, first ensure you have installed the skill via the command line: clawhub install openclaw/skills/skills/sxu75374/videochat-withme. Once installed, you must perform a one-time configuration. Run the setup script provided in the skill directory: bash skills/videochat-withme/scripts/setup.sh. This script manages your macOS launchd services, SSL certificates for secure local communication, and dependency checks. Ensure that your OpenClaw gateway is active and that the chatCompletions endpoint is enabled in your ~/.openclaw/openclaw.json configuration file. Finally, provide your Groq API key by placing it in ~/.openclaw/secrets/groq_api_key.txt so the agent can accurately transcribe your speech.

Use Cases

This skill is perfect for users seeking a more immersive AI experience. Use it to conduct "face-to-face" brainstorming sessions where the agent can visualize your whiteboard sketches via the camera, perform remote companionship, or act as a real-time tutor. It is specifically designed to handle both local desktop users (via macOS notifications) and remote users (via network IP or Tailscale), making it versatile for home office setups or mobile access when you are away from your main machine.

Example Prompts

1. "Hey, let's start a video chat so I can show you the document I'm working on."
2. "I'm heading out for lunch, can you call me on my phone so we can finish our talk?"
3. "Open up a voice call, I want to practice my presentation skills with you."

Tips & Limitations

Because this skill relies on cloud APIs for processing, please note that camera frames and audio data will be transmitted to external providers (Groq and your chosen LLM endpoint) for inference. Always check your data privacy settings within your OpenClaw configuration. For the best performance, ensure your network has sufficient upload bandwidth to stream video frames consistently. If you encounter issues with certificates, remember that your browser may require you to explicitly trust the self-signed certificate upon the first connection. If you are using a mobile device, we highly recommend using Tailscale to bypass local network restrictions and ensure a stable connection to your desktop host.