Make faceless shorts and edit video with AI
Skills used in this guide (3)
将任意文本（文章、笔记、主题、简报）转为无脸解说视频：没有可抓取的站点或素材，因此每个场景的视觉都是现编的（排版、抽象图形、图示、数据可视化）。用于主题解说、概念拆解、操作教程、清单体内容。不是由网站构建的视频（/product-launch-video——宣传或导览）。不明确时转 /hyperframes。
npx skills add heygen-com/hyperframes -g用定时、设计好的图形叠加卡——动态标题、下三分之一、数据标注、引语、侧栏、画中画——将现有口播、访谈或播客视频打包，同步到转录文本，画布可选 16:9、9:16、4:5；原片段在下方原样播放。触发词：图形叠加、屏幕图形、打包或打扮我的视频。不是纯字幕（/embedded-captions）。不明确时转 /hyperframes。
npx skills add heygen-com/hyperframes -g内嵌字幕工作流的叠加准则——字幕 MODEL（drop、rail、embed）以及字幕是合成在影片之上的 OVERLAY，而非为避让它而把内容上移预留的底部条带。在为对谈头部或发布视频添加字幕与副标题、判断某短语应 drop、沿逐字 rail 行进，或提升为稀缺的嵌入式高潮，在承载字幕的构图布局时（勿预留避让带），或在字幕下将构图定位于真实画面中心时使用。引用 embedded-captions 的 rail+embed 模型与 product-launch-video 场景 agent 的约束 #13（字幕叠加，废弃避让带）。叠加于 embedded-captions 之上。
npx skills add heygen-com/hyperframes -g
Video's pain is "shoot for hours, edit longer". The point of AI: give script and angle, let it do shot structure and captions; you only check "does it sound like me", not stare at an empty timeline.
Who this is for / not for
For:
- Faceless / knowledge / promo shorts
- No on-camera or no time to shoot
- Batch same-style clips
Not for:
- Cinematic color grading (that's an editor)
- Real news footage (AI gen needs labeling)
- No script then expect a film (script is base)
Before you start
- Script: line-by-line voiceover, clear point.
- Ratio: Douyin/Video 9:16, Bilibili 16:9.
- Assets (opt): product shots, screenrec.
Steps
1. Install the skill
faceless-explainer is a video skill. Copy its folder into your client's skills directory (same as any Agent Skill):
| Client | Directory |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex | ~/.codex/skills/ |
| Cursor | ~/.cursor/skills/ |
| Trae | ~/.trae/skills/ |
| CodeBuddy | ~/.codebuddy/skills/ |
| Tongyi Lingma | ~/.lingma/skills/ |
2. Script, then structure
Turn this voiceover into a 60s faceless explainer (9:16):
per-line visual prompt (flat vector / screenrec / icon), beat-synced, end with follow CTA.
Give shot table first, then video.3. Real footage with talking-head-recut
Own footage? talking-head-recut extracts highlights, cuts filler:
From 20min interview, pull 3 highlights of 30s, stitch by point, add transitions.4. Captions with captions-overlay
Use captions-overlay for captions, fix typos before publish.
Pitfalls
| Pit | Symptom | Fix |
|---|---|---|
| No script | Scattered | Write line-by-line first |
| Robo voice | Low retention | You read it aloud |
| Wrong ratio | Cropped | 9:16 / 16:9 first |
| Typo caption | Awkward | Proof before publish |
| Wrong platform | Throttled | Match length |
FAQ
No on-camera?
Yes, faceless-explainer is built for it.
Real footage?
talking-head-recut pulls highlights.
Captions matter?
Yes, captions-overlay boosts retention.
Ratio?
Douyin/Video 9:16, Bilibili 16:9.
Label AI?
Label generated faces/news; compliance first.
Related guides
Continue across these scenes:
- Produce a Xiaohongshu post that actually gets taps
- Generate illustrations and marketing images with AI
- Make podcast scripts and voiceover with AI
相关源码
- heygen-com/hyperframes · faceless-explainer source
- heygen-com/hyperframes · talking-head-recut source
- heygen-com/hyperframes · captions-overlay source
- Agent Skill open format spec
一句话结论
结论：短视频先给逐句脚本再让 AI 出分镜和字幕，你只把关"像不像人话"，别对着空时间轴发呆。