Major Update! Codex × Editing Skills: One-Sentence Video Editing, Free and Open Source

@chengfeng240928
СПРОЩЕНА КИТАЙСЬКА22 черв. 2026 р.
290K
546
110
26
1.3K

Коротко

This guide details a major update to the chengfeng-videocut-skills project, leveraging Codex to automate video editing, storyboarding, and HTML-based animation synthesis.

I previously created an Editing Skills project with over 2,000 stars.

Agent成峰 - inline image

This time, I've integrated Codex for a major upgrade: it can handle the subsequent processes, continuing with storyboarding, animation, and synthesis.

It has already helped me produce several videos with thousands of likes.

Agent成峰 - inline image

In the past, making a video meant opening Jianying (CapCut) first.

Now, after installing the Skills, you just give the video and script to the Agent, and it can run the entire production.

How did I do it?

Step one: Have the Agent install the Editing Skills.

The prompt is simple:

Help me install or update the chengfeng-videocut-skills editing skills. The installation address is

https://github.com/Agentchengfeng/chengfeng-videocut-skills

The Agent will confirm with you first and then run the installation itself. Once installed, these editing skills can be called locally.

Agent成峰 - inline image

Generating Edited Video and Subtitles

Step two: Organize the raw talking-head footage into a basic material package.

Agent成峰 - inline image

By giving the talking-head video and script to the "Edit Talking Head" Skill, we can get the edited video and subtitle files.

The talking-head video doesn't need to be complex. I usually just read the script once. If specific operations are involved, I switch screens and record the process.

Simply enter a slash command in the project:

/EditTalkingHead + video address + script address

After receiving the command, the Agent will first generate a review page.

Agent成峰 - inline image

The Agent will first organize pauses, slips of the tongue, and retakes. All I need to do is confirm if these deletions are correct.

Once confirmed, I click "Execute Edit."

After this step, the Agent directly outputs the edited video and subtitle files. The resulting subtitles are time-aligned with the edited video.

Generating Storyboard Pages Based on Subtitles

Once the material package is ready, enter a slash command in the project to wake up the "Talking Head Finished Product" Skill.

/TalkingHeadFinishedProduct

The Agent will generate an HTML storyboard verification page.

Agent成峰 - inline image

On the left is the visual generated by the Agent; on the right are the subtitles, visual tasks, material sources, and camera movements.

This Skill actually does three things on the storyboard page:

Agent成峰 - inline image

It first splits segments according to the subtitle timeline, then selects the visual source based on the content of each segment.

For every sentence, the Agent handles whether to keep the original video, replace it with a screenshot, a product page, a result page, or create an HTML animation right here.

If you're unsatisfied with a segment, just tell Codex which part of which segment is wrong. It's very fast to modify.

For example:

Change the animation in segment 05. Have the arrow point to the title and circle the result on the right.

Codex's Computer Use can open the page, look at the visual on the left and the speech on the right, and then go back to modify the HTML visual or annotations.

My actions have shifted from "arranging storyboards myself" to "reviewing storyboards and requesting one-click modifications."

Checking the Timeline Preview

After the storyboard page is confirmed, you can enter the timeline preview.

Agent成峰 - inline image

On the left is the video preview; at the bottom is a progress bar with text descriptions; on the right is the talking-head content.

The timeline preview arranges animations according to subtitle timestamps. Wherever the subtitles go, the animation appears there, which saves much more trouble than manually aligning timelines in editing software.

What we're checking here is: do these visuals appear correctly when put back into the full video?

If there's an error, feedback doesn't need to be a long essay; just speak by segment:

01 Keep original video. 02 Image appears too early. 03 Visual is too crowded, delete the bottom two lines of text.

Only after this step is confirmed do we enter final synthesis.

Using HyperFrames for Synthesis

After the previous video preview is confirmed, you can let the Agent synthesize it.

The Agent will use HyperFrames to turn it into a renderable video animation project.

Agent成峰 - inline image

HyperFrames is very Agent-friendly because it can turn HTML animations into video. As long as the visual can be made with HTML, it can enter this synthesis workflow.

After a few minutes, you can see the final animation.

Editing Agents are Gradually Replacing Traditional Editing

In the past, video production revolved around the timeline.

Now, video production is starting to revolve around the workflow.

Editing Agents are replacing the operational layer in traditional editing, turning video production into an automated process that can be continuously reused.

Переробити в YouMind

Перетворіть одну віральну статтю на повноцінний робочий процес

Збирайте джерела, розшифровуйте патерни, створюйте матеріали, пишіть чернетки та поширюйте контент в одному AI-робочому просторі.

Дослідити YouMind
Для авторів

Перетворіть свій Markdown на охайну статтю для 𝕏

Коли ви публікуєте власні лонгріди, зображення, таблиці та блоки коду роблять форматування в 𝕏 складним. YouMind перетворює повну чернетку в Markdown на чисту статтю для 𝕏, готову до публікації.

Спробувати Markdown для 𝕏

Більше патернів для аналізу

Останні віральні статті

Переглянути більше віральних статей