CapCut Codex plugin: ChatGPT can now rough-cut and caption your video on a PC

CapCut has published an official CapCut × Codex workflow: a plugin that lets Codex in the ChatGPT desktop app assemble a first draft of your video, edit it from the chat and hand it over to CapCut for finishing (CapCut). If you record talking-head videos, this means the rough cut and the captions can now start with a sentence typed into a chat window, as long as you are sitting at a PC.

What the plugin does

CapCut’s page lists the jobs. Codex can use a “smart rough-cut capability to organize clips and build an editable first draft”. From the chat it can shorten scenes, reorder clips, replace visuals, change text timing and update music. It generates subtitles, prepares versions in other languages, adjusts caption timing and changes caption styles. It can find a CapCut template, apply it and swap in your media (CapCut).

It is not a closed box. You can open the full CapCut editor from the conversation and carry on by hand, and CapCut says the workflow is not meant to “remove your control over the finished result” (CapCut). That is the right way to think about it: a draft generator with a chat interface, not an editor that finishes the job for you.

It runs on a desk, not in your pocket

The page says the workflow is “applied for the ChatGPT desktop app on PC, rather than as a browser-only ChatGPT video editor online” (CapCut). It does not mention CapCut on iPhone or Android, and it does not mention a Mac. It also says that “availability may vary by app version, account, and region”, so check before you plan a week of edits around it.

On price, CapCut says you can “start experiencing the CapCut × Codex workflow for free” (CapCut). The page does not say which ChatGPT plan or CapCut account you need, so we have left that open.

If you shoot on your phone, this adds a step: get the takes onto a PC, prompt, review, then export. For some creators that is already the routine. For others, the whole point of filming on a phone is that the video never leaves it.

What a prompt can and cannot know about a talking head

A rough cut of a talking-head video is mostly decisions about words: which of three takes of the opening line to keep, which repeated sentence to drop, which pause is a breath and which one lands a point. A prompt such as “shorten this and remove the slow parts” passes those decisions to the tool. Sometimes that works well. Sometimes it removes the pause you wanted.

CapCut says this itself: generation “is a starting point for editing, not an automatic guarantee of a finished video” (CapCut). So plan to watch the whole draft once. Our guide on how to edit a talking-head video in passes explains what to check on that watch, and which filler words to remove and which to keep covers the bit that automated tools tend to get wrong in both directions.

The captions are the clearer win. Generating subtitles and fixing their timing is slow by hand, and a chat command that does both is a real time saver, especially if you also post a version in a second language.

Where this fits next to the tools you already use

If your videos lean on templates, effects, music and captions, and you already finish in CapCut on a computer, the plugin is a natural addition: it handles the first draft and leaves the polish to the editor you know. Our comparison of CapCut and Descript, timeline or transcript, covers the same split between prompting a tool and choosing each cut yourself.

If you would rather make the word-by-word decisions yourself on the phone, a transcript editor is the other route. slicecut, for example, shows your take as text on iPhone or Android; you strike the words you do not want and set pauses to a target length. It does not add subtitles or templates, so for a caption-heavy, effect-heavy Reel, CapCut remains the better fit. Our page on when the CapCut timeline is the wrong tool for talking heads covers the other options.

The short version

• CapCut × Codex lets Codex in the ChatGPT desktop app on PC build a rough cut, reorder and shorten scenes, generate and time captions and apply templates, then open the result in CapCut.
• CapCut’s page does not mention phones. Plan for a desktop step.
• Treat the draft as a draft: watch it once, check the cuts in the words, keep the pauses that matter.
• For captions and templates it saves real time. For choosing which sentences stay, you are still the editor.