OpenClaw Discord Voice Input Skill

English | 简体中文

Built for one very practical reason: if you use OpenClaw through Discord, typing every prompt gets old fast.

This skill fixes that. Send your OpenClaw bot a private voice message on Discord, transcribe it locally with faster-whisper, and pass the transcript straight into OpenClaw.

Most of the annoying setup pitfalls have already been worked through, and the repository has been reviewed with skill-vetter, validated as a proper skill, and smoke-tested against a temporary OpenClaw home.

This repository is also a valid OpenClaw skill repository. SKILL.md is the skill entrypoint, and the repository root is the skill root.

Why This Skill

If your goal is simply "I want to talk to OpenClaw in Discord instead of typing all the time", this is the lightweight way to do it:

• Send the bot a voice message in a Discord private chat
• Transcribe it locally
• Feed the transcript into OpenClaw
• Get the reply back in the same chat

No voice channel plumbing. No paid speech API. No need to re-discover the same setup issues from scratch.

What It Does

• Receives private voice messages sent to your Discord bot
• Becomes the single DM ingress path for Discord private chats, so voice and text do not split into separate sessions
• Transcribes audio locally with faster-whisper
• Forwards the transcript into OpenClaw
• Deploys a macOS launchd service for the voice bridge
• Adds oc-voice-* shell aliases for lifecycle management
• Disables OpenClaw's native audio attachment preflight to avoid racing the bridge
• Disables native Discord DM intake so raw voice attachments do not create duplicate sessions

Trust Signals

• skill-vetter reviewed: checked for obvious red flags, excessive scope, and suspicious install behavior
• quick_validate passed: the repository validates cleanly as an OpenClaw skill
• install smoke-tested: the installer was run against a temporary OpenClaw home to verify file generation, config patching, and launchd setup

These badges mean the repository was reviewed and smoke-tested. They do not mean this is an official OpenClaw release, or that it is guaranteed to behave the same way on every machine.

Repository Layout

SKILL.md
agents/openai.yaml
scripts/install.py
assets/runtime/discord-dm-voice/
README.md
README.zh-CN.md

Install

Install from agentskill.sh:

https://agentskill.sh/@timrchen/openclaw-discord-voice-input-skill

/learn @timrchen/openclaw-discord-voice-input-skill

Or install from the repository directly:

python3 scripts/install.py --openclaw-home ~/.openclaw

Optional flags:

• --python /abs/path/to/python3
• --model base
• --language zh
• --skip-python-install
• --skip-npm-install
• --skip-deploy

Validate

oc-voice-status
oc-voice-logs

Then send a fresh voice message to the bot in a Discord private chat.

Notes

• Target platform: macOS with launchd
• Scope: private voice messages sent to a Discord bot
• Assumption: Discord is already configured in ~/.openclaw/openclaw.json
• Positioning: community skill, not an official OpenClaw release