菜单
WHAT: TTS audio generation with frame-accurate timing for video sync WHEN: Creating voiceover audio for explainer videos, tutorials, or any video content that needs narration
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
WHEN building any customer-facing content, funnel, page, email, call script, or offer structure THEN apply the AIDA fractal at every scale — each value ladder stage has its own AIDA cycle, each piece of content has AIDA within AIDA, and the customer journey is a Hero's Journey where the customer convinces themselves to go because they already believe themselves to be the hero.
WHAT: Dan Kennedy's framework for ascending customers through the value ladder — retention through progression WHEN: When applying direct marketing ascension knowledge
WHAT: Dan Kennedy's adaptation of Dale Carnegie's "make them feel important" for direct marketing WHEN: When applying direct marketing carnegie secret knowledge
WHAT: Dan Kennedy's Direct Marketing Diet — 6-step framework for direct marketing success WHEN: When applying direct marketing diet knowledge
WHAT: Craig Proctor's 8 Breakthrough Strategies for direct marketing success WHEN: When applying direct marketing eight breakthroughs knowledge
WHAT: Dan Kennedy's Follow-Up Framework — the 4-step structured follow-up campaign, closed loop systems, and the "no-fail" follow-up strategy. From No B.S. Direct Marketing. WHEN: When applying direct marketing follow up knowledge
| name | tts-audio-gen |
| description | WHAT: TTS audio generation with frame-accurate timing for video sync WHEN: Creating voiceover audio for explainer videos, tutorials, or any video content that needs narration |
Generate TTS audio from script text files using edge-tts, with sentence-by-sentence timing data for frame-accurate video sync.
python3 /tmp/heaven_data/skills/tts-audio-gen/scripts/generate_tts.py <scene_name> [--input-dir path] [--output-dir path] [--voice en-US-GuyNeural]
output_dir/
scene1_hj/
s01.mp3, s01.txt
s02.mp3, s02.txt
timing.json # [{index, sentence, duration_sec, frames, start_frame, file}]
files.txt # ffmpeg concat list
scene1_hj.mp3 # full concatenated audio
pip install edge-tts mutagen
# ffmpeg must be installed
Replace edge-tts call with:
from elevenlabs import generate, save
audio = generate(text=sentence, voice='Adam', model='eleven_turbo_v2_5')
save(audio, str(output_path))
API key in /home/GOD/system_config.sh as ELEVENLABS_API_KEY (if set).