Voice For Openclaw Publish

MiniMax TTS skill (enhanced). Multi-agent voice support (each agent can select a unique voice written in SOUL.md), native voice message for Telegram (MP3) an...

Security Scan

The skill's code, instructions, and required credentials align with a MiniMax TTS integration; nothing requests unrelated secrets or installs arbitrary remote code.

Purpose & Capability

Name/description, required binaries (python3, ffmpeg), and required env (MINIMAX_API_KEY) match an API-driven TTS skill that can optionally send to Telegram. Optional TELEGRAM_* vars are explained as optional for message delivery.

Instruction Scope

Instructions and scripts only call the MiniMax TTS endpoint and (optionally) Telegram/Feishu delivery paths, read a local .env and write audio files under ~/.openclaw/workspace. They do export .env keys at runtime and show examples invoking 'openclaw message send' for Feishu; these are within the stated TTS/delivery scope but users should be aware audio files and credentials are read/stored locally.

Install Mechanism

No install spec or external downloads. The skill is instruction/code-only and relies on standard binaries (python3, ffmpeg) and Python requests — no arbitrary network installs or remote archives.

Credentials

Only the MiniMax API key is required; optional Telegram vars are clearly optional. No unrelated cloud credentials or broad-scoped secrets are requested.

Persistence & Privilege

always is false and the skill does not request elevated platform privileges or modify other skills. It writes audio files under the user's workspace (~/.openclaw/workspace) which is expected for generated content.

Assessment

This skill appears to do only TTS via MiniMax and optional channel delivery. Before installing: (1) keep MINIMAX_API_KEY and any TELEGRAM_BOT_TOKEN in the local .env (do not commit it); (2) review the scripts yourself (they are included) to confirm behavior; (3) note generated audio is written to ~/.openclaw/workspace/generated — delete files if they contain sensitive content; (4) if you will use Telegram sending, provide a bot token with only the required access and verify target chat IDs; (5) confirm the API endpoint (api.minimaxi.com) is acceptable for your data/privacy needs.

MiniMax TTS Plus

Multi-agent + multi-channel native voice message TTS skill.

Core Script

All operations go through tts-xiaoye.sh (TTS generation + channel delivery).

Quick Start

bash tts-xiaoye.sh "Text to speak"

Multi-Channel Usage

Channel	Command	Format	Notes
Telegram	tts-xiaoye.sh "Text"	MP3	Direct send, no transcoding
Feishu	tts-xiaoye.sh --feishu "Text"	OGG/Opus	Auto-transcode to native voice bubble
Generate only	tts-xiaoye.sh --generate-only "Text"	MP3	Generate file without sending

Send Feishu Native Voice Message (Full Flow)

OPUS=$(bash tts-xiaoye.sh --feishu "Feishu voice content" 2>/dev/null | python3 -c "import sys,json; print(json.load(sys.stdin)['audio_file'])")
openclaw message send --channel feishu --account <YOUR_ACCOUNT_ID> --target <FeishuUserID> --media "$OPUS"

Multi-Agent Voice Configuration

Each agent can choose a unique voice and write it into their SOUL.md Voice Identity section:

## Voice Identity
- TTS model: speech-2.8-hd
- TTS voice: Chinese (Mandarin)_Warm_Girl
- TTS script: scripts/tts-xiaoye.sh

Recommended voices (verified):

Voice ID	Style	Use Case
Chinese (Mandarin)_Warm_Girl	Warm Girl	Personal Assistant
female-shaonv	Sweet Girl	Default / General
female-tianmei	Sweet Female	Gentle style
male-qn-qingse	Youthful Male	Male voice scenario
Chinese (Mandarin)_Sweet_Lady	Sweet Lady	Formal场合

List Available Voices

python3 tts-xiaoye.sh --list-voices
# or directly:
python3 scripts/tts.py --list-voices

This calls the MiniMax API and prints all available voices organized by category (System Voices, Cloned Voices, Generated Voices).

Available Models

Model	Characteristic
speech-2.8-hd	Highest quality (recommended)
speech-2.8-turbo	Faster, slightly lower quality

Full Parameters

tts-xiaoye.sh --text "Text" [--voice VoiceID] [--model Model] [--caption Caption]

Technical Notes

• TTS outputs MP3 natively. Telegram sends directly via Bot API sendVoice (MP3 supported natively).
• Feishu native voice messages require OGG/Opus format. FFmpeg handles conversion (~25ms per audio, negligible).
• FFmpeg installation: brew install ffmpeg (Linuxbrew/macOS) or apt install ffmpeg (Linux).

Setup

1. Copy setup.txt to .env and fill in your credentials:

cp skills/voice-for-openclaw/setup.txt skills/voice-for-openclaw/.env
# Then edit .env with your real values

2. The script loads credentials from .env at runtime — no hardcoded tokens in scripts.

Required env vars:

Variable	Required	Description
MINIMAX_API_KEY	✅ Yes	MiniMax API secret key (from platform.minimax.io)
TELEGRAM_BOT_TOKEN	❌ No	Telegram bot token — only needed for sending
TELEGRAM_TARGET	❌ No	Telegram chat ID — only needed together with bot token

⚠️ Security note: Credentials are loaded from .env only — no tokens are hardcoded in shell scripts. The .env file is gitignored and never published.

💡 API endpoint: The TTS API uses https://api.minimaxi.com (MiniMax's official API server), which is separate from the developer portal at platform.minimax.io.

---

MiniMax TTS Plus（多语言增强版）

多 Agent + 多渠道原生语音条增强版 TTS 技能。

核心脚本

所有操作通过 tts-xiaoye.sh 完成（TTS 生成 + 渠道发送）。

快速使用

bash tts-xiaoye.sh "要转语音的文字"

多渠道用法

渠道	命令	格式	说明
Telegram	tts-xiaoye.sh "文字"	MP3	直接发送语音条，无需转码
飞书	tts-xiaoye.sh --feishu "文字"	OGG/Opus	自动转码，发原生语音条
仅生成	tts-xiaoye.sh --generate-only "文字"	MP3	只生成文件，不发送

发送飞书原生语音条（完整流程）

OPUS=$(bash tts-xiaoye.sh --feishu "飞书语音内容" 2>/dev/null | python3 -c "import sys,json; print(json.load(sys.stdin)['audio_file'])")
openclaw message send --channel feishu --account <YOUR_ACCOUNT_ID> --target <飞书用户ID> --media "$OPUS"

多 Agent 音色配置

每个 Agent 可以选择不同音色，写入各自的 SOUL.md 的 Voice Identity 节即可：

## Voice Identity
- TTS model: speech-2.8-hd
- TTS voice: Chinese (Mandarin)_Warm_Girl
- TTS script: scripts/tts-xiaoye.sh

推荐音色（已验证）：

音色ID	风格	适用场景
Chinese (Mandarin)_Warm_Girl	温暖少女	个人助理
female-shaonv	甜美少女	默认/通用
female-tianmei	甜美女性	温柔风格
male-qn-qingse	青涩青年男	男声场景
Chinese (Mandarin)_Sweet_Lady	甜美女声	正式场合

可用模型

模型	特点
speech-2.8-hd	最高质量（推荐）
speech-2.8-turbo	快速，质量略低

完整参数

tts-xiaoye.sh --text "文字" [--voice 音色ID] [--model 模型] [--caption 文字]

技术说明

• TTS 原生输出 MP3，Telegram 直接发送（Bot API sendVoice 支持 MP3）
• 飞书原生语音条需要 OGG/Opus 格式，通过 FFmpeg 转换（耗时约 25ms/音频，可忽略）
• FFmpeg 安装方式：
  • macOS/Linuxbrew: brew install ffmpeg
  • Ubuntu/Debian: apt install ffmpeg