By Ai Ti

Edited by Shen Xiao

October 24, Audrey Irmas Pavilion, Los Angeles.

At the global creators event, CapCut World LA, CapCut (the international edition of Jianying) unveiled that its video editing and generation features will soon be integrated into ChatGPT. During the official demonstration, users uploaded five football videos and made a simple request: compile them into a 12-second vertical highlight reel titled 'GOAL!'.

ChatGPT subsequently produced a draft video. Users continued to refine their requests in natural language, prompting CapCut to adjust the footage sequence, subtitles, and timeline. After completing the initial edit, the project could be transferred to the CapCut web editor for additional effects, subtitles, refinements, and publication.

The official CapCut account described this new product as an upcoming plugin for ChatGPT. The announcement video disclosed that an open beta would be available on mobile, desktop, and web platforms starting in November.

This marks CapCut's second collaboration with ChatGPT.

In July 2023, ByteDance introduced a ChatGPT plugin for CapCut. Users could simply input a sentence, and the plugin would generate a script, voiceover, background music, subtitles, and a compilation of footage, presenting the results for user preview and editing.

This iteration of the product addressed the challenge of creating a video from scratch.

It simplified video production but operated more like an automated generator: users provided a theme, and the system produced a preliminary result, with actual modifications still necessitating a return to CapCut.

The 2026 version has evolved.

Users can now directly upload their video footage and instruct the system to comprehend, filter, and organize the content. It's no longer just about generating from a blank slate but about undertaking genuine editing tasks.

The distinction between the two can be summarized as follows:

In 2023, AI assisted in generating video ideas.

In 2026, AI begins to aid in processing a collection of footage.

The former lowered the entry barrier to creation, while the latter truly integrates into the content production workflow.

Traditionally, video editing involved multiple steps.

First, open the editing software, import the footage, manually navigate the timeline, select clips, arrange the sequence, add text, match music, and finally export versions for various platforms.

While the software offered numerous features, users needed to know their locations and the sequence for task completion.

CapCut's recent demonstration indicates that users no longer need to grasp the editing software's operational logic.

They simply need to specify the desired outcome:

'Compile these clips into a 12-second vertical football highlight reel with a fast pace and add the title GOAL!'.

The dialogue model interprets the task, while CapCut retrieves footage, organizes the timeline, and generates a rough edit.

A notable shift has occurred:

Previously, AI video products focused on 'assisting in video generation.' Now, the emphasis is shifting to 'helping transform existing footage into a usable video.'

Creating a visually complete video is no longer the most daunting task. The real time-consuming aspect lies in selecting clips, arranging them, placing subtitles, ensuring appropriate pacing, and adapting content for different platforms.

CapCut's integration into ChatGPT this time targets these repetitive yet high-frequency production tasks.

CapCut's core strength extends beyond its editing functions.

It provides templates, music, effects, stickers, transitions, and a vast library of creator content. Once users enter CapCut, they often complete footage selection, video production, modifications, and publication all within the same platform.

However, the rise of generative AI is altering the sequence in which users interact with tools.

Users may first discuss topics, write scripts, and plan shots in ChatGPT before determining the necessary footage and editing methods.

If this occurs in ChatGPT, CapCut faces a genuine risk: while it may retain the most robust editing capabilities, it may not capture the user's initial creative intentions.

Users might have already finalized content planning in the chat window, with CapCut serving merely as an 'export tool' at the end.

Therefore, CapCut must position itself at the creative process's outset.

The official declaration at CapCut World clarified this positioning shift: CapCut is no longer just a video editing tool; creators can also utilize it for design, generation, direction, and collaboration.

The plugin concept is about placing CapCut's capabilities where creative intentions are formed.

Users first articulate their objectives, and CapCut then transforms these intentions into projects, footage, and timelines.

This represents a new division of labor:

- ChatGPT is responsible for comprehending goals, supplementing requirements, and organizing tasks;

- CapCut handles footage processing, timeline management, effects application, and final video production;

- Users are tasked with confirming direction, modifying results, and finalizing publication.

What CapCut gains is not just a single invocation but the entire workflow from 'having an idea' to 'producing a final video.'

From a practical work perspective, the initial tasks to be automated will not involve aesthetics and judgment but repetitive operations.

For instance:

- Selecting usable clips from extensive footage, arranging shots according to a script or theme, deleting pauses and repetitive content, converting horizontal videos to vertical formats; automatically adding subtitles and multilingual versions, batch-generating versions with varying durations and aspect ratios for different platforms, and searching the footage library for music, transitions, and effects that match the mood and pacing.

These tasks previously required editors to constantly manipulate the timeline but do not necessarily constitute the core of creation.

The truly irreplaceable aspects still include:

- Whether the final video is accurate, natural, and believable.

In other words, the value of editors will shift from 'being proficient with software buttons' to 'defining and judging results.'