audio_codebuddy/deploy-package/scripts/parse-cosyvoice-dom.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""parse-cosyvoice-dom.py — 解析阿里云 CosyVoice 音色页面 DOM,输出 JSON + 摘要"""
import re
import json
import sys
from collections import defaultdict

PATH = r"C:/Users/caoyg/AppData/Local/Temp/cosyvoice.html"
text = open(PATH, encoding='utf-8').read()

# 切分 4 个模型段
SECTIONS = [
    ('cosyvoice-v3-flash', r'cosyvoice-v3-flash音色列表(.*?)cosyvoice-v3-plus音色列表'),
    ('cosyvoice-v3-plus',  r'cosyvoice-v3-plus音色列表(.*?)(?=下一篇|cosyvoice-v2音色列表)'),
    ('cosyvoice-v2',       r'cosyvoice-v2音色列表(.*?)cosyvoice-v1音色列表'),
    ('cosyvoice-v1',       r'cosyvoice-v1音色列表(.*?)(?=上一篇|下一篇|\Z)'),
]

all_models = {}

for model, pat in SECTIONS:
    m = re.search(pat, text, re.DOTALL)
    if not m:
        continue
    section = m.group(1)
    voices = []
    positions = [mm.start() for mm in re.finditer(r'名称[：:]\s*.+?\n', section)]
    positions.append(len(section))
    for i in range(len(positions) - 1):
        chunk = section[positions[i]:positions[i+1]]
        voice = {}
        n = re.search(r'^名称[：:]\s*(.+?)\s*$', chunk, re.MULTILINE)
        if n: voice['name'] = n.group(1).strip()
        v = re.search(r'voice参数[：:]\s*(\w+)', chunk)
        if v: voice['voice_id'] = v.group(1)
        t = re.search(r'特质[：:]\s*(.+?)\n', chunk)
        if t: voice['desc'] = t.group(1).strip()[:80]
        a = re.search(r'年龄[：:]\s*(.+?)\n', chunk)
        if a: voice['age'] = a.group(1).strip()
        l = re.search(r'语言[：:]\s*(.+?)\n', chunk)
        if l: voice['lang'] = l.group(1).strip()[:60]
        voice['instruct'] = '支持' if 'Instruct：支持' in chunk else '不支持' if 'Instruct：不支持' in chunk else '?'
        # scene/地区
        s = re.search(r'适用场景[：:]\s*(.+?)\n', chunk)
        if s: voice['scene'] = s.group(1).strip()[:40]
        if voice.get('name'):
            voices.append(voice)
    all_models[model] = voices

# 输出 JSON
with open(r"C:/Users/caoyg/AppData/Local/Temp/cosyvoice-voices.json", "w", encoding="utf-8") as f:
    json.dump(all_models, f, ensure_ascii=False, indent=2)

# 输出摘要
sys.stdout.reconfigure(encoding='utf-8')
for model, voices in all_models.items():
    print(f"\n=== {model}: {len(voices)} 个音色 ===")
    lang_count = defaultdict(int)
    for v in voices:
        lang = v.get('lang', '?')
        # 简化为第一个语种
        first = lang.split('、')[0].split('（')[0]
        lang_count[first] += 1
    print(f"  语种分布:")
    for lang, cnt in sorted(lang_count.items(), key=lambda x: -x[1]):
        print(f"    {lang}: {cnt}")