← Back to Skills Marketplace
senseaudio-let-claw-talk
by
Wu Ruixiao
· GitHub ↗
· v1.0.0
· MIT-0
121
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install senseaudio-let-claw-talk
Description
当用户希望把 AudioClaw 变成一个持续监听、开口就说、停顿就回答的本机语音助手时使用。这个 skill 会在 macOS 上启动常驻监听流程，默认优先使用内置 Swift 录音器减少 Python 音频依赖；用户语音通过 SenseAudio ASR 转文字，再发给 audioclaw agent，并用...
Usage Guidance
This skill appears to do what it says: run a local, persistent macOS voice assistant using SenseAudio for ASR/TTS and an optional local WeSpeaker service. Before installing or running it, consider: 1) It will open a Terminal and start a long-lived process that captures your microphone — only enable if you trust the code and want continuous listening. 2) It will create and modify files under ~/.audioclaw/workspace/state (preferences, voice clone data, embeddings) and may start a local background Python service for speaker verification. 3) You must provide SENSEAUDIO_API_KEY; that key will be sent to https://api.senseaudio.cn for ASR/TTS — treat it as a secret and only use a key you trust to share with that service. 4) The bundle expects swiftc and python3 to be present and may compile or run the provided Swift mic capture binary. 5) If you have privacy or security concerns, review run_continuous_voice_assistant.py and the SenseAudio client files (local_senseaudio_asr.py, local_senseaudio_tts.py) yourself to confirm there are no additional network endpoints or unexpected behaviors, and verify the WeSpeaker service code before enrolling voice prints. If any of these behaviors are unacceptable (continuous mic access, storing voiceprints, sending audio to SenseAudio), do not enable the skill.
Capability Assessment
Purpose & Capability
Name and description (continuous, always-listening macOS voice assistant) match the actual files and runtime behavior. Declared requirements (python3, swiftc, SENSEAUDIO_API_KEY) are reasonable: Swift code captures microphone audio, Python implements SenseAudio ASR/TTS wrappers and a controller, and a local weSpeaker service is optional. Network calls target senseaudio.cn for ASR/TTS which is coherent with the stated SenseAudio integration.
Instruction Scope
SKILL.md tells the agent to open a Terminal and run the provided start/launch scripts to spawn a long-lived microphone process; the included scripts read/write state under ~/.audioclaw/workspace/state, invoke local binaries (/usr/bin/afplay, /usr/bin/afinfo), spawn subprocesses, and may start a local background service (WeSpeaker). All of these actions are within the expected scope for a continuous voice assistant, but they are privacy-sensitive (continuous microphone capture, local voiceprint enrollment) and the instructions deliberately start persistent processes rather than one-shot tasks.
Install Mechanism
No external install spec (no remote downloads) — the skill is instruction + local scripts. That lowers install-time risk. The only build-time requirement is swiftc (declared) to compile the provided Swift microphone capture source if needed; otherwise scripts rely on system Python and bundled Python scripts. No URLs/archives are fetched by the skill itself.
Credentials
Only one required credential is declared (SENSEAUDIO_API_KEY) and the code uses it to call SenseAudio endpoints (ASR/TTS), which is appropriate. The bundle also reads many optional VOICECLAW_/AUDIOCLAW_* environment variables and will write/read files under ~/.audioclaw/workspace/state; these are configuration options, not additional secrets. Users should note SENSEAUDIO_API_KEY will be sent to https://api.senseaudio.cn and should treat it as a secret.
Persistence & Privilege
always:false and user-invocable:true — the skill does not auto-enable itself but is explicitly designed to start long-running, always-listening processes (Terminal window, persistent microphone capture, optional WeSpeaker background service). It will create state directories and save voiceprint embeddings and other runtime state under the user's home (~/.audioclaw). This level of persistence is expected for the declared purpose but is privacy-sensitive.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install senseaudio-let-claw-talk - After installation, invoke the skill by name or use
/senseaudio-let-claw-talk - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
senseaudio-let-claw-talk 1.0.0
- Initial release: enables a persistent SenseAudio voice assistant on macOS with default settings for hands-free, multi-turn conversations.
- Minimizes audio backend dependencies by preferring the built-in Swift recorder over Python audio packages.
- Integrates SenseAudio ASR for speech recognition, audioclaw agent for dialogue, and SenseAudio TTS or system "say" for responses.
- Default setup does not require WeSpeaker speaker verification or streaming TTS, but both remain optional for advanced users.
- Provides flexible voice preferences, interrupter sensitivity, and “wake/sleep” phrase logic designed for natural, on-device voice assistant use.
- Emphasizes robust user feedback, persistent settings, and compatibility with existing AudioClaw/OpenClaw workflows.
Metadata
Frequently Asked Questions
What is senseaudio-let-claw-talk?
当用户希望把 AudioClaw 变成一个持续监听、开口就说、停顿就回答的本机语音助手时使用。这个 skill 会在 macOS 上启动常驻监听流程，默认优先使用内置 Swift 录音器减少 Python 音频依赖；用户语音通过 SenseAudio ASR 转文字，再发给 audioclaw agent，并用... It is an AI Agent Skill for Claude Code / OpenClaw, with 121 downloads so far.
How do I install senseaudio-let-claw-talk?
Run "/install senseaudio-let-claw-talk" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is senseaudio-let-claw-talk free?
Yes, senseaudio-let-claw-talk is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does senseaudio-let-claw-talk support?
senseaudio-let-claw-talk is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created senseaudio-let-claw-talk?
It is built and maintained by Wu Ruixiao (@kikidouloveme79); the current version is v1.0.0.
More Skills