Initial commit: English Wonder World

給 Jay 的小朋友英文學習網頁，包含 Units 1-4、端午節特輯、
數字單元。含 Gemini TTS 預錄音檔與 LINE Seed 字型。

generate_audio.py

@@ -0,0 +1,146 @@
+#!/usr/bin/env python3
+"""
+generate_audio.py — 用 OpenAI TTS 把 english_units.html 裡所有英文句子跑成 mp3。
+
+用法：
+    export OPENAI_API_KEY=你的key
+    python3 generate_audio.py [--force] [--voice shimmer] [--speed 0.9]
+
+不需要任何套件（純 Python 標準函式庫），不需要 ffmpeg（OpenAI 直接回 mp3）。
+
+聲音：alloy / ash / ballad / coral / echo / fable / nova / onyx / sage / shimmer / verse
+模型：gpt-4o-mini-tts（預設，支援 instructions）/ tts-1-hd / tts-1
+"""
+import argparse
+import hashlib
+import json
+import os
+import re
+import sys
+import time
+import urllib.error
+import urllib.request
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent
+HTML = ROOT / "english_units.html"
+AUDIO_DIR = ROOT / "audio"
+OPENAI_TTS_URL = "https://api.openai.com/v1/audio/speech"
+
+INSTRUCTIONS = (
+    "Speak in a warm, playful, encouraging tone like a cheerful older sister reading a "
+    "picture book to a young child. Clear American English, slightly slower than normal, "
+    "full of warmth and excitement. Sound friendly and approachable, not stiff. "
+    "Pronounce every word clearly."
+)
+
+
+def extract_texts(html_text: str):
+    pattern = re.compile(r'en:\s*"((?:[^"\\]|\\.)*)"')
+    clean, seen = [], set()
+    for m in pattern.findall(html_text):
+        t = m.replace('\\"', '"').replace("\\'", "'")
+        t = re.sub(r'<[^>]+>', '', t).strip()
+        if t and t not in seen:
+            seen.add(t)
+            clean.append(t)
+    return clean
+
+
+def hash_text(text: str) -> str:
+    return hashlib.sha1(text.encode('utf-8')).hexdigest()[:16]
+
+
+def tts(api_key: str, text: str, voice: str, model: str,
+        fmt: str = 'mp3', speed: float = 0.9) -> bytes:
+    payload = {
+        "model": model,
+        "input": text,
+        "voice": voice,
+        "response_format": fmt,
+        "speed": speed,
+    }
+    if model.startswith("gpt-4o"):
+        payload["instructions"] = INSTRUCTIONS
+    req = urllib.request.Request(
+        OPENAI_TTS_URL,
+        data=json.dumps(payload).encode('utf-8'),
+        headers={
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        },
+    )
+    with urllib.request.urlopen(req, timeout=90) as r:
+        return r.read()
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument('--force', action='store_true', help='重新產生已存在的檔案')
+    ap.add_argument('--voice', default='nova',
+                    help='alloy/ash/ballad/coral/echo/fable/nova/onyx/sage/shimmer/verse')
+    ap.add_argument('--speed', type=float, default=0.92,
+                    help='0.25–4.0，<1 變慢（預設 0.92）')
+    ap.add_argument('--model', default='gpt-4o-mini-tts',
+                    choices=['gpt-4o-mini-tts', 'tts-1-hd', 'tts-1'])
+    ap.add_argument('--format', default='mp3',
+                    choices=['mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'])
+    ap.add_argument('--sleep', type=float, default=0.2)
+    args = ap.parse_args()
+
+    api_key = os.getenv('OPENAI_API_KEY')
+    if not api_key:
+        print('需要 OPENAI_API_KEY', file=sys.stderr)
+        print('  去 https://platform.openai.com/api-keys 拿', file=sys.stderr)
+        sys.exit(1)
+
+    AUDIO_DIR.mkdir(exist_ok=True)
+    texts = extract_texts(HTML.read_text(encoding='utf-8'))
+    print(f'找到 {len(texts)} 句不重複英文')
+    print(f'模型 {args.model}　聲音 {args.voice}　速度 {args.speed}x　格式 {args.format}')
+    print()
+
+    manifest = {}
+    made = skipped = failed = 0
+
+    for i, text in enumerate(texts, 1):
+        h = hash_text(text)
+        out = AUDIO_DIR / f'{h}.{args.format}'
+        label = text if len(text) <= 55 else text[:52] + '...'
+
+        if out.exists() and not args.force:
+            manifest[text] = out.name
+            skipped += 1
+            print(f'[{i:3}/{len(texts)}] skip   {label}')
+            continue
+
+        print(f'[{i:3}/{len(texts)}] gen    {label}')
+        try:
+            audio = tts(api_key, text, args.voice, args.model, args.format, args.speed)
+            out.write_bytes(audio)
+            manifest[text] = out.name
+            made += 1
+        except urllib.error.HTTPError as e:
+            failed += 1
+            body = e.read().decode('utf-8', errors='ignore')[:240]
+            print(f'          HTTP {e.code}: {body}')
+        except Exception as e:
+            failed += 1
+            print(f'          FAIL: {e}')
+
+        time.sleep(args.sleep)
+
+    (AUDIO_DIR / 'manifest.json').write_text(
+        json.dumps(manifest, ensure_ascii=False, indent=2), encoding='utf-8')
+    (AUDIO_DIR / 'manifest.js').write_text(
+        'window.__audioManifest = ' + json.dumps(manifest, ensure_ascii=False) + ';\n',
+        encoding='utf-8',
+    )
+
+    print()
+    print(f'完成：新產生 {made}、跳過 {skipped}、失敗 {failed}')
+    print(f'檔案在 {AUDIO_DIR}/')
+
+
+if __name__ == '__main__':
+    main()