🦞 Claw Body — Give Your Claw a Body

Every Claw deserves a body.

Turn your OpenClaw AI into a real-time digital avatar — with a face, a voice, and expressions. Talk to your Claw face-to-face, not just through cold text.

NEW: Presentation Mode 🎤 — Your Claw can now be a presenter! Load a PPT/PDF and let the avatar narrate your slides.

Free 5-minute trial included. Sign up at nuwaai.com to create your own custom avatar for free.

For Every Claw Fan

• - 🎨 Design your dream Claw — cute, anime, realistic, handsome, beautiful, or buff — your call
• 🗣️ Voice chat — speak to your Claw and hear it talk back with lip-sync
• 📺 Real-time video — see your Claw's expressions as it responds
• 🧠 Same brain — it's your OpenClaw agent, just with a face. Same memory, same personality
• 🌐 中文 / English — bilingual interface with language toggle
• 📊 Presentation mode — narrate PPT/PDF slides with digital avatar (works with claw-presenter skill)

Quick Start

When user runs /claw-body:

1. 1. Start the server:

1. 2. Tell the user:

How It Works

CODEBLOCK1

This skill uses NuwaAI's humanctrl mode with ASR:

• - Your voice → NuwaAI speech recognition → text
• Text → OpenClaw Gateway → agent generates reply
• Reply → drives the avatar's voice and lip movements

Same agent, new interface. The avatar is just another channel — like iMessage or Telegram, but with a face.

Features

• - 🎤 Real-time voice input (ASR)
• 🗣️ Lip-synced avatar speech
• 🧠 Same OpenClaw agent — not a separate bot
• 📺 WebRTC real-time video stream
• 💬 Text input fallback
• 📱 Auto-adapts to portrait / landscape / square avatars
• 🔧 In-browser config — zero env vars needed
• 🎁 Free 5-min trial with demo avatar
• 🌐 Chinese / English bilingual UI
• 🔄 Disconnect / reconnect controls

Create Your Own Avatar

1. 1. Go to nuwaai.com — sign up is free
2. Create your avatar — first one is free!
3. Get your API Key, Avatar ID, and User ID
4. Enter them in the Claw Body interface
5. Done — your Claw now has a body 🦞

Requirements

• - OpenClaw Gateway running
• NuwaAI account (free sign-up)
• Modern browser (WebRTC + microphone)
• Node.js 18+

⚠️ When User Asks to Present PPT/PDF

If the user says anything like "讲PPT"、"讲解PDF"、"帮我讲解演示文件" while in Claw Body:

DO NOT open or operate any application (Keynote, PowerPoint, Preview) on the user's computer.

Instead, follow this flow:

1. 1. Ask for the file path if not provided
2. Run the Claw Presenter parse script:

   python3 <workspace>/skills/claw-presenter/scripts/parse-presentation.py "<file-path>"
   

1. 3. Read the generated INLINECODE1
2. For slides without scripts (script is empty), generate narration based on content
3. Update presentation.json with the generated scripts
4. Tell the user the presentation is ready and ask to start
5. When user confirms, reply with [PRESENTATION_START:<output-dir>] to enter presentation mode
6. Narrate slide by slide using [SLIDE:N] tags

📊 Presentation Mode (Slide Playback)

Claw Body supports a presentation mode for narrating pre-prepared slide decks.

Prerequisites

Use the Claw Presenter skill first to prepare a presentation folder:
CODEBLOCK3

Each slide in presentation.json has an image path and a script (narration text).

Entering Presentation Mode

When user wants to present a prepared folder, include this tag in your reply:
CODEBLOCK4

Narrating Slides

For each page, read the script field from presentation.json and reply with:
CODEBLOCK5

The frontend automatically flips to the corresponding slide image.

User Controls During Narration

• - "继续" / "下一页" → advance to next slide
• "上一页" → go back
• "跳到第N页" → jump
• Any question → answer it, then user says "继续" to resume
• "退出讲解" → reply with INLINECODE11

Control Tags

Important

• - Do NOT operate PPT/PDF applications on the user's computer — slides are shown in the browser
• Read narration from presentation.json script field — don't regenerate unless asked
• Keep a natural speaking pace, one slide at a time

⚠️ First-Time Setup (Important!)

Claw Body talks to your OpenClaw agent via the Gateway's chat API. This endpoint is disabled by default and must be enabled once.

When user runs /claw-body:

1. 1. Read INLINECODE18
2. If gateway.http.endpoints.chatCompletions.enabled is not true, tell the user to enable it manually:

   > openclaw gateway restart
   > 

Add this to ~/.openclaw/openclaw.json:

CODEBLOCK7