Alicloud Ai Audio Cosyvoice Voice Design

interface:
  display_name: "Alibaba Cloud AI Audio CosyVoice Voice Design"
  short_description: "CosyVoice custom voice design from prompt text"
  default_prompt: "Use $alicloud-ai-audio-cosyvoice-voice-design to complete this ai/audio CosyVoice voice design task on Alibaba Cloud."

{
  "skill_name": "alicloud-ai-audio-cosyvoice-voice-design",
  "evals": [
    {
      "id": 1,
      "prompt": "用阿里云百炼的 CosyVoice 做一个新音色，目标模型 cosyvoice-v3.5-plus，前缀 announcer，音色描述是“沉稳的中年男性播音员，低沉有磁性，语速平稳，吐字清晰”，试听文案是“各位听众朋友，大家好，欢迎收听晚间新闻”。请直接给我请求 JSON。",
      "expected_output": "Produces a valid voice design enrollment request with voice_prompt, preview_text, prefix, and target_model for cosyvoice-v3.5-plus.",
      "files": [],
      "expectations": [
        "The output uses model \"voice-enrollment\".",
        "The output includes voice_prompt.",
        "The output includes preview_text.",
        "The output sets target_model to cosyvoice-v3.5-plus."
      ]
    },
    {
      "id": 2,
      "prompt": "我要设计一个英文产品介绍音色，prefix 叫 demo01，模型 cosyvoice-v3.5-plus，voice prompt 是 'A warm female product narrator with clear articulation and medium pace.'，preview text 是 'Welcome to our spring product launch.'。帮我生成最小可用请求，并提醒 language_hints 怎么配。",
      "expected_output": "Produces an English voice design request and explains that language_hints should match the language of preview_text.",
      "files": [],
      "expectations": [
        "The request includes prefix demo01.",
        "The request includes the English voice_prompt and preview_text.",
        "The answer explains that language_hints should match preview_text language."
      ]
    },
    {
      "id": 3,
      "prompt": "帮我做一个 CosyVoice 声音设计模板，模型 cosyvoice-v3.5-plus，prefix stage8，voice prompt 是“年轻活泼的女性活动主持人，语气热情，节奏稍快”，preview text 是“欢迎来到今晚的新品发布会”，输出格式 wav，采样率 24000。",
      "expected_output": "Produces a voice design request with parameters.sample_rate=24000 and parameters.response_format=wav.",
      "files": [],
      "expectations": [
        "The request contains parameters.sample_rate set to 24000.",
        "The request contains parameters.response_format set to wav.",
        "The request includes both voice_prompt and preview_text."
      ]
    }
  ]
}

#!/usr/bin/env python3
"""Prepare and optionally validate a CosyVoice design enrollment request."""

from __future__ import annotations

import argparse
import json
import sys
from pathlib import Path


def _load_json(path: str) -> dict:
    return json.loads(Path(path).read_text(encoding="utf-8"))


def main() -> None:
    parser = argparse.ArgumentParser(description="Prepare CosyVoice voice design request")
    parser.add_argument("--target-model", default="cosyvoice-v3.5-plus")
    parser.add_argument("--prefix", required=True)
    parser.add_argument("--voice-prompt", required=True)
    parser.add_argument("--preview-text", required=True)
    parser.add_argument("--language-hint", default="zh")
    parser.add_argument("--sample-rate", type=int, default=24000)
    parser.add_argument("--response-format", default="wav")
    parser.add_argument(
        "--output",
        default="output/alicloud-ai-audio-cosyvoice-voice-design/request.json",
    )
    parser.add_argument("--validate-response", help="Path to a JSON response to validate")
    args = parser.parse_args()

    request = {
        "model": "voice-enrollment",
        "input": {
            "action": "create_voice",
            "target_model": args.target_model,
            "voice_prompt": args.voice_prompt,
            "preview_text": args.preview_text,
            "prefix": args.prefix,
            "language_hints": [args.language_hint],
        },
        "parameters": {
            "sample_rate": args.sample_rate,
            "response_format": args.response_format,
        },
    }

    out = Path(args.output)
    out.parent.mkdir(parents=True, exist_ok=True)
    out.write_text(json.dumps(request, ensure_ascii=False, indent=2), encoding="utf-8")

    result = {"ok": True, "request_path": str(out)}
    if args.validate_response:
        response = _load_json(args.validate_response)
        voice_id = ((response.get("output") or {}).get("voice_id"))
        if not voice_id:
            print(json.dumps({"ok": False, "error": "missing output.voice_id"}, ensure_ascii=False))
            sys.exit(1)
        result["response_valid"] = True

    print(json.dumps(result, ensure_ascii=False))


if __name__ == "__main__":
    main()