openclaw-yandex-speechkit

Yandex SpeechKit speech synthesis provider for OpenClaw.

Adds Yandex SpeechKit as a TTS provider, enabling voice replies in Telegram, Talk Mode, and any channel that supports voice messages. Uses the SpeechKit REST API v3 with OGG/Opus output for compact, Telegram-native voice bubbles.

Install

openclaw plugins install openclaw-yandex-speechkit

Or install manually and register in your config:

npm install openclaw-yandex-speechkit

Configuration

Set your Yandex Cloud credentials:

# In .openclaw/.env
YANDEX_API_KEY=your-api-key
YANDEX_FOLDER_ID=your-folder-id

Then configure TTS in openclaw.json:

{
  "tts": {
    "provider": "yandex"
  },
  "plugins": {
    "entries": {
      "openclaw-yandex-speechkit": {
        "enabled": true,
        "config": {
          "apiKey": "${YANDEX_API_KEY}",
          "folderId": "${YANDEX_FOLDER_ID}",
          "voice": "alexander",
          "role": "good",
          "speed": 1.0
        }
      }
    }
  }
}

Authentication

Two auth modes are supported:

| Mode | Env var | Header | Use case | |------|---------|--------|----------| | API Key | YANDEX_API_KEY | Api-Key <key> | Service accounts, long-lived | | IAM Token | YANDEX_IAM_TOKEN | Bearer <token> | Short-lived (12h), federated accounts |

The plugin auto-detects IAM tokens (they start with t1.). You can also set authType explicitly in config.

When using IAM tokens, folderId is required.

Voices

16 voices across 5 languages:

| Voice | Gender | Language | Roles | |-------|--------|----------|-------| | alexander | Male | Russian | good, evil, neutral | | ermil | Male | Russian | good, neutral | | zahar | Male | Russian | good, evil | | filipp | Male | Russian | — | | madirus | Male | Russian | — | | marina | Female | Russian | friendly, whisper, neutral | | oksana | Female | Russian | friendly, strict, neutral | | jane | Female | Russian | good, evil, neutral | | omazh | Female | Russian | good, evil, neutral | | masha | Female | Russian | good, strict, friendly | | dasha | Female | Russian | friendly, good, neutral | | amira | Female | Kazakh | — | | madi | Male | Kazakh | — | | john | Male | English | good, neutral | | lea | Female | German | — | | nigora | Female | Uzbek | — |

See the full voice reference for details.

Audio formats

| Format | Config value | Extension | Best for | |--------|-------------|-----------|----------| | OGG/Opus | oggopus | .opus | Telegram voice bubbles (default) | | MP3 | mp3 | .mp3 | General playback | | WAV | wav | .wav | Lossless, post-processing |

The plugin automatically selects OGG/Opus for voice-note targets (Telegram, Discord) and uses the configured format for other contexts.

Config reference

All fields under plugins.entries.openclaw-yandex-speechkit.config:

| Field | Type | Default | Description | |-------|------|---------|-------------| | apiKey | string | $YANDEX_API_KEY | API key or IAM token | | folderId | string | $YANDEX_FOLDER_ID | Yandex Cloud folder ID | | voice | string | ermil | Voice name | | role | string | — | Voice role (emotion): good, evil, friendly, strict, neutral, whisper | | speed | number | 1.0 | Speaking speed (0.1 - 3.0) | | format | string | oggopus | Output format: oggopus, mp3, wav | | authType | string | auto | Auth mode: apiKey or iamToken (auto-detected from token prefix) |

Development

git clone [email protected]:mrkhachaturov/openclaw-yandex-speechkit.git
cd openclaw-yandex-speechkit
npm install
npm run build
npm test

Local development with OpenClaw

Install as a path-based plugin:

// .openclaw/config/plugins.json5
{
  "entries": {
    "yandex-speechkit": { "enabled": true }
  },
  "installs": {
    "yandex-speechkit": {
      "source": "path",
      "installPath": "/path/to/openclaw-yandex-speechkit"
    }
  }
}

License

MIT