Edit videos in CapCut with AI: just ask Denker
Give it the raw recording and say what you want. Denker cuts it, adds subtitles and builds the CapCut project.
Summary
Denker is a desktop AI agent that edits your videos in CapCut for you. You give it a video and a goal. It transcribes the footage on your computer, cuts it, adds a title and subtitles, and checks its own preview. Then you get an editable CapCut project, a finished file, or both. The edit runs in a background helper, so Denker stays free to talk, and it asks you before anything is posted.
Key points
- Say what you want in plain words: a vertical short, a shorter cut, captions, a title.
- Give it an example video, and Denker copies its style: cut rhythm, captions and zooms.
- Denker can fill a YouTube, TikTok, Instagram or LinkedIn upload page, then waits for your yes before it publishes.
You recorded a product demo. It is long, it has pauses, and it is horizontal. You want a vertical short with subtitles for TikTok. Now you can edit videos in CapCut with AI: you hand the recording to Denker, say what you want, and Denker does the edit in CapCut and tells you when it is done.
Watch: Denker fixes a shot in a vertical TikTok project in CapCut
Why does a raw recording never turn into a short?
Because the edit is the slow part. You scrub the timeline to find the good bits. You cut the pauses and the false starts. You reframe for vertical. Then you type or fix every caption by hand.
How many recordings are sitting on your desktop right now, waiting for an evening you never have?
Denker takes that work off your hands. You stay the editor in chief: you say what you want, and you check the result.
How to edit videos in CapCut with AI, step by step
1. Give Denker the video. Drop the file, give its path, or just say “my last screen recording”. If two files match, Denker asks which one.
2. Say what you want. For example: “Make a 60-second vertical short for TikTok with big captions.” Denker asks only what you left open, like the platform, the length, or whether you want a CapCut project or a finished file.
3. Denker hands the edit to one background helper. Denker stays free to talk, asks you anything it needs, and tells you when the edit is done.
4. The helper transcribes the footage, plans the cuts, the title and the subtitles, and renders a preview. It watches that preview itself and fixes what is off before you see it.
5. You get the result. Denker tells you what changed, such as the length before and after and the number of cuts, and where each file is.
Turn a long video into shorts in a style you like
Have a creator whose shorts you like? Give Denker one of their videos as a file. Denker measures the cut rhythm and looks at the captions, zooms and title. Then it describes the style in plain words and asks you to correct it.
Denker keeps the style in your project. Next time, you skip the example and just say which style to use.
Add subtitles to video automatically
Denker transcribes the speech on your computer, with a dictation model you already have that hears the video's language. Only when none does, it asks you to download the Standard model (Whisper) in Settings. The video tools it needs are fetched once, the first time.
Subtitles are placed on the edited timeline, so they stay in sync with the cuts. Denker also reads them back to catch wrong words before you see the video.
AI agent video editing inside CapCut itself
Denker can write an editable CapCut project, open it in CapCut and keep working there. CapCut has few controls an agent can read, so Denker reads the window from a screenshot and uses the real pointer. It does the clicks and drags it planned in one go: pick a tab, drag a clip onto the timeline, click a field and type. When something on screen changed, it stops and looks again.
Using the pointer needs your consent. You can see what Denker does and stop it.
That is what the video above shows: Denker selects a clip, changes its scale so the shot fills the vertical frame, pastes text and plays the preview. It reports while it works. When you want the file, Denker can press Export in CapCut, or render the finished file itself.
The 3D model shot is fixed. The weird glow is off and the whole engine with its labels is centered. Shots 3 to 5 still need the same check.
You decide what gets posted
When you ask, Denker opens the YouTube, TikTok, Instagram or LinkedIn upload page in your own browser and fills it in. Then it stops and asks for your clear yes before Publish or Post. It never posts on its own.
Your original video is never changed. Denker never signs in to CapCut or a platform for you, never buys CapCut Pro, and never changes your own CapCut projects. It also won't download a video from a link: platform terms forbid that, so it asks you for the file.
- CapCut installed on your Mac or Windows PC
- Claude Code, Codex or OpenCode, signed in with your own account
- The video as a file on your computer
One desktop AI agent for video and everything else
Denker is not a separate video app. It is the desktop AI agent you already use for email, sheets and code, on Mac and Windows, with the Claude Code, Codex or OpenCode you already pay for. Video editing is one more thing you can ask it.
The video style Denker learns is saved in your project, next to your Docs and tasks, so your next edit starts from the same AI agent workspace.
CapCut is made by another company. Denker is not affiliated with it; it works with the CapCut app you have installed.
Frequently asked questions
Do I need CapCut?
Yes. CapCut must be installed on your Mac or Windows PC. Then you choose an editable CapCut project, a finished file, or both.
Does Denker post my video without asking?
No. Denker fills the upload page in your browser, then asks for your clear yes before it presses Publish or Post.
Can Denker edit a video from a YouTube or TikTok link?
No. Platform terms forbid downloading those videos, so Denker asks you for the file instead.
Which AI do I need?
Claude Code, Codex or OpenCode, with your own account. Charges from your AI provider are separate.