Interactive English listening tutor using audio transcription (faster-whisper) and text-to-speech (edge-tts). Use when user sends English audio for learning, wants to practice listening comprehension, do dictation exercises, or study from podcasts/audiobooks/BBC content. Triggers on phrases like "学英语", "练听力", "听这段", "英语学习", "next segment", "下一段", or when user sends audio files asking for transcription/practice.
用户发音频 → whisper 转写 → 分段播放(edge-tts) → 用户复述 → 批改反馈 → 下一段
当用户发送音频文件(mp3/m4a/wav 等):
python3 ~/.agents/skills/english-tutor/scripts/transcribe.py <audio_path> [--lang en] [--model tiny]
en(英文)tiny(快,适合长音频);可选 small/medium(更准)python3 ~/.agents/skills/english-tutor/scripts/speak.py "<text>" --voice en-US-GuyNeural --rate -10% --output /tmp/english_tutor_segment.mp3
message tool 发送音频到当前频道见 references/voices.md — 按场景选声音:
en-US-GuyNeural(默认,自然亲切)en-US-JennyNeuralen-GB-RyanNeuralen-GB-SoniaNeural语速建议:
-10% 或 -5%(稍慢)+0%+10%用户复述后,对照原文做三档评价:
| 档位 | 标准 | 示例反馈 |
|---|---|---|
| ✅ 核心信息全对 | 主要事实、数字、人名正确 | "完全正确!补充:xxx" |
| ⚠️ 大意对但缺细节 | 抓住了主旨但漏了关键细节 | "方向对!还提到了:xxx" |
| ❌ 偏离较大 | 遗漏主要信息或理解错误 | "再听一遍?关键点是:xxx" |
反馈原则:
用户说"词汇总结"或"vocabulary"时,输出本课所有重点词汇表:
建议从 -10% 开始,用户适应后逐步提升到 +0% 甚至 +10%。
用户可以发任何英文音频(播客、新闻、演讲、视频截图),不限于 BBC。
pip install faster-whisper edge-tts{{OPENCLAW_MEDIA_DIR}}/inbound/<uuid>.<ext>/tmp/ 或 workspace temp_ 前缀,用完可清理