minimax-speech

MiniMax 语音合成技能 - 支持同步/异步文本转语音(T2S)、音色克隆(Voice Clone)、音色设计(Voice Design)、音色查询与删除。使用模型 speech-2.8-hd，输出 mp3/wav/pcm 格式音频文件到本地。

Security Scan

Code and docs implement a MiniMax TTS/voice-clone client (coherent with the description) but the package metadata omits required credentials and dependency declarations and the source is unknown — proceed cautiously.

Purpose & Capability

The skill's code and SKILL.md implement TTS, voice cloning, voice design, list/get/delete voice operations using a MiniMax HTTP API — that matches the described purpose. However, the registry metadata lists no required environment variables or dependencies, while both SKILL.md and scripts/speech.py require MINIMAX_API_KEY (and optionally MINIMAX_REGION) and the Python 'requests' library. The missing manifest declarations and unknown source/homepage are an incoherence and reduce traceability.

Instruction Scope

Runtime instructions and the script stay within the stated purpose: they call MiniMax API endpoints, write output audio files locally, and read reference audio files for cloning. The SKILL.md does instruct providing MINIMAX_API_KEY and MINIMAX_REGION (which is consistent with the code). There are no instructions to read unrelated user files or system secrets beyond the API key or to send data to unexpected endpoints. Note: cloning uploads base64-encoded audio to the third-party API (privacy/legal implication).

Install Mechanism

This is instruction-only with a bundled Python script (no install spec). That is lower risk, but the script depends on the 'requests' package which is not declared in metadata. There is no automated installer; users will need to ensure Python and requests are present. No arbitrary downloads or extract/install steps are present.

Credentials

The functionality legitimately requires a MINIMAX_API_KEY and optionally MINIMAX_REGION; these are referenced in SKILL.md and enforced by the code. However the registry metadata claims 'Required env vars: none', which is inconsistent and misleading. No other credentials are requested. The script reads local audio files when cloning (expected) and writes output files — this is proportional but users should be aware that audio data is uploaded to the remote service.

Persistence & Privilege

The skill does not request always:true, does not modify other skills or system configurations, and does not persist credentials itself. Autonomous invocation is allowed by default (normal), but there is no elevated persistence or privileged behavior in the package.

What to consider before installing

This package appears to implement the advertised TTS and voice-clone features, but the manifest is incomplete and the source/origin is unknown. Before installing: 1) Do not use a production/broadly-permissioned API key — create a restricted/test key or billing limits. 2) Verify the API domains (api.minimaxi.com / minimax.io) and the provider's legitimacy; ask the author for a homepage or repository. 3) Ensure Python and the 'requests' library are available; the manifest should declare this dependency. 4) Understand privacy: cloning uploads your reference audio (potentially sensitive) to a third-party service — get consent from speakers. 5) Prefer running first in an isolated environment (container or VM) to observe network traffic and behavior. 6) Ask the publisher to correct the registry metadata to declare MINIMAX_API_KEY and MINIMAX_REGION and to provide verifiable source code hosting; absence of these is the main reason this skill is flagged as suspicious.

MiniMax Speech Skill

使用 MiniMax API 进行语音合成、语音克隆和音色设计。支持同步/异步文本转语音、音色克隆、音色设计、音色查询与删除。使用模型 speech-2.8-hd，输出 mp3/wav/pcm 格式音频文件。

环境配置

{
  "MINIMAX_API_KEY": "your-api-key",
  "MINIMAX_REGION": "cn" | "int"
}

• MINIMAX_API_KEY: MiniMax API 密钥
• MINIMAX_REGION: 区域设置，cn 为中国，int 为国际（默认 cn）

可用函数

text_to_speech(text, voice_id, output_file)

同步文本转语音

参数:

• text: 要转换的文本
• voice_id: 音色ID（默认: female-tianmei）
• output_file: 输出文件路径

示例: text_to_speech("你好世界", "female-tianmei", "hello.mp3")

text_to_speech_async(text, voice_id)

异步文本转语音，返回任务ID

参数:

• text: 要转换的文本
• voice_id: 音色ID

返回: 任务ID

query_speech_task(task_id)

查询异步任务状态

参数:

• task_id: 任务ID

返回: 任务状态信息

clone_voice(audio_file_path, title)

音色快速复刻

参数:

• audio_file_path: 参考音频文件路径
• title: 新音色名称

返回: 新音色的 voice_id

design_voice(text, style)

音色设计

参数:

• text: 音色描述文本
• style: 音色风格

返回: 设计生成的 voice_id

list_voices()

获取音色列表

返回: 音色列表

get_voice(voice_id)

获取单个音色信息

参数:

• voice_id: 音色ID

delete_voice(voice_id)

删除音色

参数:

• voice_id: 音色ID

返回: 是否成功

常用音色

音色ID	描述
female-tianmei	女声甜美
male-yunyang	男声播音
female-badu	女声巴度
male-shawn	男声 Shawn
female-shanshan	女声杉杉

使用示例

同步语音合成

python scripts/speech.py tts "欢迎使用 MiniMax 语音服务" -v female-tianmei -o output.mp3

异步语音合成

python scripts/speech.py tts-async "这是一段较长的文本" -v male-yunyang

查询任务

python scripts/speech.py query <task_id>

克隆音色

python scripts/speech.py clone reference_audio.mp3 -t "我的音色"

音色设计

python scripts/speech.py design "温柔的女性声音，适合朗读" -s gentle

删除音色

python scripts/speech.py delete <voice_id>

注意事项

1. 同步语音合成适合短文本（< 60秒音频）
2. 长文本建议使用异步接口
3. 音色克隆需要清晰、无噪音的参考音频
4. 克隆和设计的音色需要审核后才能使用