openclaw-volcengine-tts

Volcengine (Doubao) speech provider plugin for OpenClaw — high-quality Chinese TTS powered by ByteDance's speech synthesis engine.

Features

• 🎙️ First-class speech provider — works with messages.tts (auto TTS) and Talk Mode (real-time voice conversation)
• 🇨🇳 Optimized for Chinese — native Mandarin voices with natural prosody
• 🌍 Multi-language — also supports English, Japanese, Korean, and multilingual voices
• 😊 Emotion control — happy, sad, angry, and more (on supported voices)
• ⚡ Low latency — direct API calls, no wrapper overhead
• 🔧 Fully configurable — voice, speed, pitch, volume, encoding, cluster

Quick start

1. Install

npm install -g openclaw-volcengine-tts

2. Set credentials

Get your App ID and Access Token from Volcengine Speech Console.

export VOLCENGINE_TTS_APP_ID="your-app-id"
export VOLCENGINE_TTS_ACCESS_TOKEN="your-access-token"

Or add to ~/.openclaw/.env:

VOLCENGINE_TTS_APP_ID=your-app-id
VOLCENGINE_TTS_ACCESS_TOKEN=your-access-token

3. Configure OpenClaw

For message auto-TTS (replies sent as voice):

// ~/.openclaw/openclaw.json
{
  messages: {
    tts: {
      auto: "always",
      provider: "volcengine-tts",
    },
  },
}

For Talk Mode (real-time voice conversation on phone/node):

{
  talk: {
    provider: "volcengine-tts",
    providers: {
      "volcengine-tts": {
        voiceId: "zh_female_yingtaowanzi_mars_bigtts",
      },
    },
  },
}

4. Restart OpenClaw

openclaw gateway restart

Configuration

Plugin config (optional)

{
  plugins: {
    entries: {
      "volcengine-tts": {
        config: {
          voiceType: "zh_female_yingtaowanzi_mars_bigtts",
          cluster: "volcano_tts",
          encoding: "mp3",        // mp3 | pcm | ogg_opus | wav
          speedRatio: 1.0,        // 0.5 - 2.0
          volumeRatio: 1.0,       // 0.5 - 2.0
          pitchRatio: 1.0,        // 0.5 - 2.0
          emotion: "happy",       // optional, on supported voices
          language: "zh",         // optional, explicit language hint
        },
      },
    },
  },
}

Environment variables

| Variable | Description | |----------|-------------| | VOLCENGINE_TTS_APP_ID | App ID from Volcengine console | | VOLCENGINE_TTS_ACCESS_TOKEN | Access token | | VOLCENGINE_TTS_VOICE | Default voice type (fallback) |

Available voices

Run /voice list in OpenClaw to see all available voices. Some highlights:

Chinese Female

• zh_female_yingtaowanzi_mars_bigtts — 樱桃丸子 (warm & sweet)
• zh_female_shuangkuaisisi_moon_bigtts — 爽快思思 (crisp & clear)
• zh_female_sajiaonvyou_moon_bigtts — 撒娇女友 (cute)
• zh_female_gaolengyujie_moon_bigtts — 高冷御姐 (cool & mature)

Chinese Male

• zh_male_chunhou_moon_bigtts — 醇厚 (deep & mature)
• zh_male_yangguang_moon_bigtts — 阳光 (bright & positive)

Multilingual

• multi_female_shuangkuaisisi_moon_bigtts — supports CN/EN/JP/KR
• multi_male_jingqiangkanye_moon_bigtts — supports CN/EN/JP/KR

English

• en_female_amanda_mars_bigtts — Amanda
• en_male_adam_mars_bigtts — Adam

Full voice list: Volcengine Voice Catalog

Talk Mode usage

Once configured, Talk Mode works on any OpenClaw node device (iOS/Android/macOS):

1. Pair your phone with your OpenClaw gateway
2. Enable Voice Wake (e.g., wake word "管家")
3. Start talking — responses come back in the configured Chinese voice

This turns any old phone into a Chinese-speaking AI assistant.

How it works

The plugin calls the Volcengine OpenSpeech TTS API directly via HTTP. No SDK dependency, no WebSocket complexity — just a clean REST call that returns base64-encoded audio.

Pricing

Volcengine TTS pricing is usage-based. See Volcengine Pricing for current rates. Generally very affordable for personal use.

License

MIT