小米 MiMo TTS 接入 OpenClaw 问题复盘暨 QQBot 语音三层兜底方案#

1
cd /home/node/.openclaw/workspace/skills/xiaomi-mimo-v2-tts/scripts && python3 tts_compat.py --port 18888 &

1
{
2
  "models": {
3
    "providers": {
4
      "openai": {
5
        "baseUrl": "http://localhost:18888/v1",
6
        "apiKey": "dummy",
7
        "models": []
8
      }
9
    }
10
  },
11
  "messages": {
12
    "tts": {
13
      "auto": "always",
14
      "provider": "openai"
15
    }
16
  }
17
}

1
if (voiceError === "Voice generate failed" && text?.trim()) {
2
    // 第二层：尝试手动 TTS
3
    const manualResult = await manualTtsFallback(text);
4
    if (manualResult.success && manualResult.audioPath) {
5
        // 用手动生成的音频发送
6
        const result = await sendVoice(target, manualResult.audioPath, ...);
7
        if (!result.error) return result;
8
    }
9
    // 第三层：都失败，直接发文字
10
    await sendTextAfterMedia(target, text);
11
    return { error: "manual_tts_failed" };
12
}

1
async function manualTtsFallback(text) {
2
    const timestamp = Date.now();
3
    const outputPath = `/tmp/manual_tts_${timestamp}.mp3`;
4
    const scriptPath = "/home/node/.openclaw/workspace/skills/xiaomi-mimo-v2-tts/scripts/tts.py";
5

6
    // 调用 Python 脚本生成 TTS
7
    execSync(`python3 "${scriptPath}" --text "${text}" --output "${outputPath}" --format mp3`);
8

9
    // 检查文件是否生成成功
10
    if (fs.statSync(outputPath).size > 0) {
11
        return { success: true, audioPath: outputPath };
12
    }
13
    return { success: false };
14
}

1
services:
2
  openclaw:
3
    env_file:
4
      - .env
5
    command: >
6
      sh -c "python3 /path/to/tts_compat.py --port 18888 &
7
             node dist/index.js gateway --bind lan --port 18789"

1
[Unit]
2
Description=OpenClaw MiMo TTS Compatibility Layer
3
After=network.target
4

5
[Service]
6
Type=simple
7
ExecStart=/usr/bin/python3 /path/to/tts_compat.py --port 18888
8
Restart=always
9

10
[Install]
11
WantedBy=multi-user.target