🔥 Prometheus Avatar — OpenClaw Plugin

Give your OpenClaw agent a Live2D avatar with real-time lip-sync, emotion, TTS — plus AAA image generation (gpt-image-1 backed) for marketplace skin previews.

🚀 Install

openclaw plugins install @prometheusavatar/openclaw-plugin

✨ What it does

This plugin bridges OpenClaw agent events to the Prometheus Avatar SDK, giving your AI agent a visual body and the ability to create marketplace assets:

• 🎭 Live2D Avatar — Renders a Live2D model in your agent's UI
• 🗣️ Text-to-Speech — Agent messages spoken aloud with lip-sync
• 😊 Emotion Detection — Text sentiment drives expressions (happy / sad / angry / surprised / thinking)
• 🎨 AAA Image Generation (NEW v0.9) — Generate game-store-tier skin preview cards directly from your agent conversation, backed by OpenAI GPT Image 2 (gpt-image-1)
• 🛒 Marketplace Asset Pipeline — Generate thumbnails + deploy assets without leaving the agent loop

⚙️ Configuration

Add to your openclaw.config.json:

{
  "plugins": {
    "prometheus-avatar": {
      "avatarId": "haru",
      "ttsProvider": "web-speech",
      "enableLipSync": true,
      "enableEmotion": true
    }
  }
}

| Option | Type | Default | Description | |--------|------|---------|-------------| | avatarId | string | — | Avatar model ID from Prometheus Marketplace | | modelUrl | string | Haru (default) | Direct URL to a .model3.json file | | ttsProvider | string | "web-speech" | TTS provider: web-speech / elevenlabs / azure / sherpa-onnx | | ttsVoice | string | — | Voice name or ID | | enableLipSync | boolean | true | Audio-driven lip synchronization | | enableEmotion | boolean | true | Emotion analysis from text |

🛠️ Agent Tools (3)

When your OpenClaw agent has tool-use enabled, the plugin registers 3 creator tools:

| Tool | Description | |------|-------------| | prometheus_generate_image_pro | NEW v0.9 Generate AAA-quality images via gpt-image-1 (skin preview cards, posters, UI mocks). Genshin / Overwatch / WoW shop card tier. 9 style presets · BYOK OpenAI · Free quota · Pro Credits. | | prometheus_generate_thumbnail | Generate marketplace asset thumbnails (legacy FLUX route — kept for backward compat) | | prometheus_deploy_asset | Deploy new assets (voices, skins, effects) to the Marketplace |

Example: Generate a skin preview card

User: "Generate a cyberpunk anime girl skin in AAA Genshin Impact shop card tier"

Agent (via prometheus_generate_image_pro):
  {
    style: "cyberpunk",
    taskType: "aaa_skin",
    size: "1024x1536",
    quality: "high",
    prompt: "3D cel-shaded engine render, anime girl with neon hair and tech visor, slight elevated 3/4 hero pose, glossy game-art materials, clean dark studio backdrop with subtle radial gradient, cyan / magenta rim lighting, NOT flat 2D illustration, Genshin Impact / Overwatch shop preview tier, production-ready AAA skin preview card."
  }

Returns 1024×1536 base64 image (or publicUrl when upload: true) — ready to ship to the marketplace.

💡 Twin Prompt-Is-The-Ceiling rule: Recommend ≥100-word prompts with explicit AAA benchmark named (Genshin / Overwatch / WoW / Pixar / Studio Ghibli). Lazy short prompts produce mediocre output — the model can render anything if you describe it precisely.

📡 Events

| Listens to | Emits | |------------|-------| | agent:message → speaks + animates | avatar:speak | | agent:thinking → thinking expression | avatar:emotion | | agent:error → surprised expression | avatar:ready |

🌍 Ecosystem

• OpenClaw — Merged into the official community plugin registry (4/21/2026, by Peter Steinberger)
• Hermes Agent — Plugin PR in queue at NousResearch/hermes-agent#9773
• Cursor / Claude Code / Any MCP Client — Use the @prometheusavatar/mcp-server (9 tools) for direct MCP access to the same image engine

🔗 Links

• SDK: @prometheusavatar/core v0.11+
• MCP server: @prometheusavatar/mcp-server v0.3+
• Marketplace: prometheus.mythslabs.ai
• Forge UI (visual reference): marketplace/create — 7-style picker
• GitHub: myths-labs/prometheus-avatar

📄 License

MIT © Myths Labs