kClawd

AI voice assistant, image generation, and utility tools — voice conversations via Discord, image creation via OpenRouter, currency/commodity conversion.

Modules

• voice-bot — Discord voice bot. Joins channels, listens to speech with wake word detection, responds via TTS. TypeScript, discord.js, faster-whisper, Piper TTS.
• generators/image-gen — Text-to-image and image-to-image generation via OpenRouter (Gemini 3 Pro). 1K/2K/4K resolution.
• generators/converter — Currency and cryptocurrency converter with commodity equivalents. Supports 190+ fiat currencies, 19 cryptos, metals, and fun items. German language support.

Architecture

┌──────────────┐     ┌──────────────────┐     ┌───────────────┐
│  Discord VC  │────▶│   voice-bot      │────▶│  OpenClaw     │
│  (Opus)      │◀────│  STT → LLM → TTS │◀────│  Gateway      │
└──────────────┘     └──────────────────┘     └───────────────┘

┌──────────────┐     ┌──────────────────┐     ┌───────────────┐
│  User/Agent  │────▶│   image-gen      │────▶│  OpenRouter   │
│  (CLI)       │◀────│  prompt → image  │◀────│  (Gemini 3)   │
└──────────────┘     └──────────────────┘     └───────────────┘

┌──────────────┐     ┌──────────────────┐     ┌───────────────┐
│  User/Agent  │────▶│   converter      │────▶│  Public APIs  │
│  (CLI/HTTP)  │◀────│  currency + item │◀────│  (Free tier)  │
└──────────────┘     └──────────────────┘     └───────────────┘

Voice Pipeline

1. Capture — Audio from Discord (Opus 48kHz stereo)
2. STT — faster-whisper (GPU, large-v3) transcribes speech
3. Wake word — Detects "bot"/"god"/"pot" + variants, filters filler words
4. LLM — OpenClaw Gateway generates a response
5. TTS — Piper TTS (local, CPU-based, German/English)
6. Playback — Audio streamed back to the voice channel

Requirements

• Linux (WSL2 supported)
• Node.js 22+ / Bun (for converter)
• Python 3.10+ with faster-whisper and piper-tts
• NVIDIA GPU with CUDA 12.x (for voice bot Whisper acceleration)
• ffmpeg
• OpenClaw Gateway running on port 18789 (for voice bot)
• OpenRouter API key (for image generation)

Quick Start

Voice Bot (Discord)

cd voice-bot
cp .env.example .env   # fill in Discord token + Gateway token
npm install
./start.sh

Commands: !join, !leave, !say <text>, !status, !help

See voice-bot/README.md for full setup.

Image Generation

cd generators/image-gen
# Set OPENROUTER_API_KEY in .env
./imggen "a cyberpunk city at night"
./imggen "make the sky sunset" -i photo.jpg -r 2k

See generators/image-gen/README.md for details.

Currency Converter

cd generators/converter
bun converter.ts "15.5 DOT" "HKD" --lang de
bun converter.ts "100 EUR" "USD" --commodity gold
bun converter.ts --serve --port 3456  # HTTP API mode

Supports crypto (BTC, ETH, DOT, etc.), fiat currencies, and commodity equivalents (metals, liquids, fun items).

See generators/converter/README.md for details.

Project Structure

kClawd/
├── voice-bot/              # Voice bot workspace
│   ├── voice-bot/          # Application source (TypeScript)
│   │   └── src/            # index.ts, voice-receiver.ts, etc.
│   ├── shared/             # Shared modules (STT, TTS, Gateway client, logger)
│   ├── start.sh            # Entry point with prerequisite checks
│   └── .env                # Configuration
├── generators/
│   ├── image-gen/          # Image generation tool (OpenRouter)
│   └── converter/          # Currency + commodity converter (Bun)
├── shared/                 # Root-level shared modules (reference copy)
├── venv-piper/             # Python venv for Piper TTS
└── tsconfig.base.json      # Shared TypeScript config

HTTP APIs

Voice Bot

• POST http://localhost:18790/speak — Speak text through active voice channel
• GET http://localhost:18790/health — Health check + active sessions
• GET http://localhost:18790/status — Uptime and session count

Converter

• GET http://localhost:3456/convert?amount=20&from=PLN&to=USD — Convert currency
• GET http://localhost:3456/commodities — List all commodities
• GET http://localhost:3456/currencies?base=USD — List available currencies
• GET http://localhost:3456/health — Health check

Features

Voice Bot

• Wake word detection with fuzzy matching
• GPU-accelerated speech transcription (faster-whisper)
• Multilingual TTS (Piper: German + English)
• Automatic silence detection and utterance boundaries
• Conversation logging to transcriptions.jsonl
• HTTP API for external control

Image Generation

• Text-to-image and image-to-image editing
• Multi-resolution support (1K, 2K, 4K)
• Auto-naming from prompts
• Cost tracking (USD per request)
• JSON output with file paths

Converter

• 190+ fiat currencies + 19 cryptocurrencies
• Live precious/industrial metal pricing
• Dynamic liquid+container commodities (glass of beer, bathtub of oil, etc.)
• Silly abstract items (hamburgers, therapy sessions, parking tickets)
• German language support (--lang de)
• 24h cache with daily warmup (cron at 00:30)
• HTTP API mode

License

Internal project.