← 返回 Skills 市场

Free voice from Comfy UI + Qwen3 TTS

Name: Free voice from Comfy UI + Qwen3 TTS
Author: onimka

作者 Onimka · GitHub ↗ · v1.0.0

cross-platform ⚠ suspicious

2252

总下载

当前安装

版本数

在 OpenClaw 中安装

/install free-voice

功能描述

Generate Russian male voice audio using ComfyUI with Qwen3 TTS node and save as MP3 for voice messages.

使用说明 (SKILL.md)

TTS через ComfyUI

Описание

Навык для генерации аудио через ComfyUI с использованием узла AILab_Qwen3TTSVoiceDesign_Advanced.

Требования

ComfyUI с установленным плагином AILab_Qwen3TTSVoiceDesign_Advanced
Модель Qwen3
Папка E:\Ai\Comfy UI\output\ для сохранения аудио

Запуск ComfyUI

Если ComfyUI не запущена, использовать:

cmd /c start "" "C:\Users\user\Desktop\ComfyUI.lnk"

Ожидание готовности

Проверять доступность по http://localhost:8000 каждые 10 секунд до готовности.

Формат запроса

{
  "prompt": {
    "50": {
      "inputs": {
        "filename_prefix": "qwen3-tts/[UNIQUE_ID]",
        "quality": "320k",
        "audioUI": "",
        "audio": ["55", 0]
      },
      "class_type": "SaveAudioMP3",
      "_meta": {"title": "Сохранить аудио (MP3)"}
    },
    "55": {
      "inputs": {
        "text": "[TEXT]",
        "instruct": "A male voice with a slightly hoarse, warm tone, speaking in a confident and friendly manner.",
        "model_size": "1.7B",
        "device": "auto",
        "precision": "bf16",
        "language": "Russian",
        "max_new_tokens": 2048,
        "do_sample": true,
        "top_p": 0.9,
        "top_k": 50,
        "temperature": 0.9,
        "repetition_penalty": 1,
        "unload_models": false,
        "seed": -1
      },
      "class_type": "AILab_Qwen3TTSVoiceDesign_Advanced",
      "_meta": {"title": "Qwen3 TTS VoiceDesign (Advanced)"}
    }
  }
}

Обработка

Отправить POST-запрос на http://localhost:8000/prompt
Получить prompt_id
Ждать завершения в /history/[prompt_id]
Найти файл в E:\Ai\Comfy UI\output\[filename_prefix]*.mp3
Отправить как голосовое сообщение через message(asVoice=true)

安全使用建议

This skill is coherent for local TTS generation, but check these before installing: 1) Ensure you actually have ComfyUI, the AILab_Qwen3TTS plugin, and the Qwen3 model installed and licensed as required. 2) Confirm the hard-coded paths/shortcut match your machine (the SKILL.md uses C:\Users\user\Desktop\ComfyUI.lnk and E:\Ai\Comfy UI\output\); if not, edit the instructions or don't install. 3) Be aware the skill will try to start a local application and read a local folder (it needs permission to launch processes and access the filesystem). 4) The skill only talks to localhost, but generated audio may be sent out by the agent via your messaging function (message(asVoice=true))—consider whether that's acceptable for the content you plan to synthesize. 5) If you run this in an environment without those local resources, the skill will fail rather than silently do something else.

功能分析

Type: OpenClaw Skill Name: free-voice Version: 1.0.0 The skill is classified as suspicious due to the direct command execution instruction `cmd /c start "" "C:\Users\user\Desktop\ComfyUI.lnk"` found in `SKILL.md`. While this command is plausibly intended to launch a local application (ComfyUI) required for the skill's function, direct shell command execution by an AI agent represents a significant security risk. Additionally, the skill instructs the agent to access the local file system to retrieve generated audio from `E:\Ai\Comfy UI\output\`, which, while necessary for its purpose, contributes to the overall elevated risk profile.

能力评估

✓ Purpose & Capability

The skill's name (ComfyUI + Qwen3 TTS) matches the instructions: it needs ComfyUI, the AILab_Qwen3TTSVoiceDesign_Advanced node, the Qwen3 model, and an output folder to save mp3 files. Nothing requested appears unrelated to the stated TTS purpose.

ℹ Instruction Scope

Instructions are narrowly scoped: start ComfyUI (via a desktop shortcut command), poll http://localhost:8000, POST a JSON prompt, wait for completion, then read the generated MP3 from a local output folder and send it as a voice message. Minor concerns: the instructions use hard-coded, system-specific paths (C:\Users\user\Desktop\ComfyUI.lnk and E:\Ai\Comfy UI\output\) which will fail or behave unexpectedly on other machines; the skill will attempt to launch a local shortcut and read a local folder, so it requires the agent/runtime to have access to launch processes and read that path.

✓ Install Mechanism

No install spec or downloaded code; this is instruction-only, so nothing is written to disk by the skill itself and no external packages are fetched.

✓ Credentials

No environment variables, credentials, or config paths are requested. The only resources referenced are local ComfyUI service on localhost and a local output directory—these are proportional to generating and retrieving TTS output.

✓ Persistence & Privilege

always is false and the skill does not request persistent system-wide changes or cross-skill configuration. It will launch a local process (ComfyUI) and access local files, which is expected for this functionality.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install free-voice
安装完成后，直接呼叫该 Skill 的名称或使用 /free-voice 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

- Initial release of TTS skill for generating audio via ComfyUI using the AILab_Qwen3TTSVoiceDesign_Advanced node. - Requires pre-configured ComfyUI with the specified plugin and Qwen3 model. - Defines detailed audio generation and saving steps, including request/response workflow. - Output audio is saved to a specific folder and delivered as a voice message. - Instruction for automatic ComfyUI startup and server readiness monitoring included.

元数据

Slug free-voice

版本 1.0.0

许可证 —

累计安装 8

当前安装数 8

历史版本数 1

常见问题