# Faceless-explainer Skill
**作者**: HeyGen
**日期**: 2026-07-10T20:42:32.000Z
**来源**: [https://x.com/HeyGen/status/2075682104302882908](https://x.com/HeyGen/status/2075682104302882908)
---
# Day 5 of 30: Turn any topic into an Explainer video with zero footage
Today the input is just a concept. No website to capture, no footage, no assets.
/faceless-explainer turns text (a topic, an article, your notes) into a finished explainer video where every visual is invented: kinetic typography, diagrams, data cards, count-ups. The skill plans a teaching story frame by frame, writes and voices the narration, builds each frame as HTML in your brand system, adds captions, music, and sound effects, then renders the MP4. One prompt in, one video out.
## One prompt, one FRAME.md:
We gave it the FRAME.md we built for heygen.com (the Day 2 discipline: our colors, our fonts, our rules), dropped a song of ours into the project, and asked for a 45-second explainer.
The topic: Agent Skills themselves.
prompt:
```
Use HyperFrames /faceless-explainer and create a 45-second
video explaining skills for agents:
1. First, explain what skills are, why they're useful,
and how they work.
2. Then find a popular skill on GitHub and choose one strong example.
3. Learn how it works and explain what the skill does, why you
picked it, and why it's useful.
4. Use our FRAME.md provided, (our HeyGen brand system)
for the design system.
Make it accurate, clear, and compelling.
```
## What the agent actually did
1. Researched for real. It hit the GitHub API, found the most-starred skills repo (Superpowers, 250,924 stars), read the systematic-debugging skill inside it, and built the script from what it learned. Every number on screen is live data, not vibes.
2. Invented every visual. A folder that assembles itself, a SKILL.md anatomy diagram, a star counter, a four-phase ladder: all designed from our FRAME.md, our fonts, our blue.
3. Finished the film. Voiceover with word timings, captions, our song trimmed and faded to length, sound effects on the cuts, lint and validation checks, rendered to MP4.
Initial output:
this was a one shot. We asked for an explainer video, and that was it. Besides the prompt, the FRAME.md, and the song we dropped in, the agent had full creative direction: the story, the script, the research, the visuals, the pacing. Every call you see was its own.
## It looked good. We changed one thing:
We did not like the caption style. The word being spoken sat inside an animated pill, and it fought the frames. So we asked to change the caption style and remove the pill around the word being said. Now it is simple color: grey for what is coming, blue on the word being spoken.
One sentence, not a paragraph. The agent restyled the captions, re-checked the video, and rendered again.
Final output:
Everything you just watched came out of HyperFrames from one prompt: the research, the script, the voiceover, the motion graphics, the captions, the brand system. We never opened an editor.
Enjoying the series?
⭐ HyperFrames on GitHub: https://github.com/heygen-com/hyperframes
## 相关链接
- [HeyGen](https://x.com/HeyGen)
- [@HeyGen](https://x.com/HeyGen)
- [9.1K](https://x.com/HeyGen/status/2075682104302882908/analytics)
- [heygen.com](http://heygen.com/)
- [https://github.com/heygen-com/hyperframes](https://github.com/heygen-com/hyperframes)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [4:42 AM · Jul 11, 2026](https://x.com/HeyGen/status/2075682104302882908)
- [9,196 Views](https://x.com/HeyGen/status/2075682104302882908/analytics)
- [View quotes](https://x.com/HeyGen/status/2075682104302882908/quotes)
---
*导出时间: 2026/7/11 13:01:23*
---
## 中文翻译
# Faceless-explainer 技能
**作者**: HeyGen
**日期**: 2026-07-10T20:42:32.000Z
**来源**: [https://x.com/HeyGen/status/2075682104302882908](https://x.com/HeyGen/status/2075682104302882908)
---
# 30天系列第5天：将任何主题转化为无需实拍的解说视频
今天的输入仅仅是一个概念。没有需要抓取的网站，没有素材画面，也没有现成的资源。
/faceless-explainer 能将文本（一个主题、一篇文章、你的笔记）转化成一部完成的解说视频，其中的每一个视觉元素都是凭空生成的：动态排版、图表、数据卡片、数字增长动画。该技能会逐帧规划教学故事，撰写并朗读旁白，按照你的品牌系统构建每一帧 HTML，添加字幕、音乐和音效，最后渲染成 MP4。一个提示词输入，一个视频输出。
## 一个提示词，一个 FRAME.md：
我们将为 heygen.com 构建的 FRAME.md（即第 2 天的规范：我们的颜色、字体和规则）放入项目，加入了一首我们的歌曲，并要求生成一个 45 秒的解说视频。
主题：Agent 技能本身。
提示词：
```
使用 HyperFrames /faceless-explainer 创建一个 45 秒的视频，讲解 Agent 技能：
1. 首先，解释什么是技能，为什么它们有用，以及它们是如何工作的。
2. 然后在 GitHub 上找到一个流行的技能库，并挑选一个强有力的示例。
3. 学习它的工作原理，解释该技能的作用、你选择它的原因以及它的用处。
4. 使用提供的 FRAME.md（我们的 HeyGen 品牌系统）作为设计系统。
确保内容准确、清晰且引人入胜。
```
## Agent 实际做了什么
1. **真枪实弹地调研。** 它调用了 GitHub API，找到了星标最多的技能库（Superpowers，250,924 星），阅读了其中的 systematic-debugging 技能，并基于所学内容构建了脚本。屏幕上显示的每一个数字都是实时数据，并非凭空臆造。
2. **发明了每一个视觉画面。** 一个自动组装的文件夹，一张 SKILL.md 结构图解，一个星标计数器，一个四阶段阶梯图：所有设计都基于我们的 FRAME.md、我们的字体、我们的蓝色。
3. **完成了影片。** 包含单词时机的配音、字幕，经过修剪并淡入淡出以适应长度的歌曲，转场处的音效，以及代码检查和验证步骤，最终渲染为 MP4。
初始输出：
这是一次性生成的。我们要了一个解说视频，仅此而已。除了提示词、FRAME.md 和我们放入的歌曲外，Agent 拥有完全的创作主导权：故事、脚本、调研、视觉、节奏。你看到的每一次调用都是它独立完成的。
## 看起来不错。我们改了一处：
我们不喜欢字幕的样式。正在朗读的单词被包裹在一个动态的药丸形状中，这与画面冲突了。所以我们要求更改字幕样式，并移除朗读单词周围的胶囊框。现在颜色更简洁：即将朗读的词显示为灰色，正在朗读的词显示为蓝色。
只需一句话，而不是一段话。Agent 重新设计了字幕样式，重新检查了视频，并再次进行了渲染。
最终输出：
你刚才看到的所有内容——调研、脚本、配音、动态图形、字幕、品牌系统——都来自 HyperFrames 的一个提示词。我们从未打开任何编辑器。
喜欢这个系列吗？
⭐ GitHub 上的 HyperFrames: https://github.com/heygen-com/hyperframes
## 相关链接
- [HeyGen](https://x.com/HeyGen)
- [@HeyGen](https://x.com/HeyGen)
- [9.1K](https://x.com/HeyGen/status/2075682104302882908/analytics)
- [heygen.com](http://heygen.com/)
- [https://github.com/heygen-com/hyperframes](https://github.com/heygen-com/hyperframes)
- [升级到 Premium](https://x.com/i/premium_sign_up)
- [2026年7月11日 上午4:42](https://x.com/HeyGen/status/2075682104302882908)
- [9,196 次观看](https://x.com/HeyGen/status/2075682104302882908/analytics)
- [查看引用](https://x.com/HeyGen/status/2075682104302882908/quotes)
---
*导出时间: 2026/7/11 13:01:23*