L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

tts-whatsapp

tts-whatsapp

Send high-quality text-to-speech voice messages on WhatsApp in 40+ languages with automatic delivery

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: tts-whatsapp version: 1.0.0 description: Send high-quality text-to-speech voice messages on WhatsApp in 40+ languages with automatic delivery user-invocable: true disable-model-invocation: false tags:


🎙️ TTS WhatsApp - Voice Messages in 40+ Languages

Send high-quality text-to-speech voice messages on WhatsApp with automatic delivery. Supports 40+ languages, personal messages, and group broadcasts.

✨ Features

  • 🎙️ High-quality TTS powered by Piper (40+ languages)
  • 🎵 Automatic conversion to OGG/Opus (WhatsApp format)
  • 📤 Automatic sending via Clawdbot
  • 👥 Group support - Send to individuals or WhatsApp groups
  • 🌍 Multi-language - French, English, Spanish, German, and 40+ more
  • 🧹 Smart cleanup - Auto-delete files after successful send
  • ⚡ Fast - ~2-3s from command to delivery

📦 Prerequisites

  1. Piper TTS: pip3 install --user piper-tts
  2. FFmpeg: brew install ffmpeg (macOS) or apt install ffmpeg (Linux)
  3. Voice models: Download from Hugging Face
    • Place in ~/.clawdbot/skills/piper-tts/models/
    • Example: fr_FR-siwis-medium.onnx

🚀 Quick Start

Basic usage

tts-whatsapp "Hello, this is a test" --target "+15555550123"

Send to WhatsApp group

tts-whatsapp "Hello everyone" --target "120363257357161211@g.us"

Change language

tts-whatsapp "Hola mundo" --lang es_ES --voice carlfm --target "+34..."

Different quality levels

tts-whatsapp "High quality" --quality high --target "+1..."

🌍 Supported Languages

  • 🇫🇷 French (fr_FR): siwis, upmc, tom
  • 🇬🇧 English GB (en_GB): alan, alba
  • 🇺🇸 English US (en_US): lessac, amy, joe
  • 🇪🇸 Spanish (es_ES, es_MX): carlfm, davefx
  • 🇩🇪 German (de_DE): thorsten, eva_k
  • 🇮🇹 Italian (it_IT): riccardo
  • 🇵🇹 Portuguese (pt_BR, pt_PT): faber
  • 🇳🇱 Dutch (nl_NL): mls, rdh
  • 🇷🇺 Russian (ru_RU): dmitri, irina
  • And 30+ more!

Full voice list →

🔧 Configuration

Configure in ~/.clawdbot/clawdbot.json:

{
  "skills": {
    "entries": {
      "tts_whatsapp": {
        "enabled": true,
        "env": {
          "WHATSAPP_DEFAULT_TARGET": "+15555550123",
          "PIPER_DEFAULT_LANG": "en_US",
          "PIPER_DEFAULT_VOICE": "lessac",
          "PIPER_DEFAULT_QUALITY": "medium"
        }
      }
    }
  }
}

🎛️ All Options

--target NUMBER       WhatsApp number or group ID
--message TEXT        Text message with audio
--lang LANGUAGE       Language (default: fr_FR)
--voice VOICE         Voice name (default: auto)
--quality QUALITY     x_low, low, medium, high
--speed SPEED         Playback speed (default: 1.0)
--no-send            Don't send automatically

📊 Performance

~2.3s total for a 10-second message:

  • TTS generation: ~1s
  • Format conversion: ~0.2s
  • WhatsApp delivery: ~1s

📚 Full Documentation

See README.md for complete documentation, examples, and troubleshooting.

Skills enregistrées