Can ChatGPT edit videos? What it can and cannot do

Can ChatGPT edit videos?

Not in the sense an editor means: it does not take your recording and hand back a finished, cut video. What it can do with a video file is limited, and OpenAI says so plainly. As of 18 October 2026, OpenAI's help centre says:

ChatGPT accepts video files as attachments, including on Free plans, with availability varying by platform, upload method and account.
It may use tools to analyse an uploaded video, but that analysis can be incomplete or inaccurate. It may not analyse the entire video or accurately interpret its audio.
Video attachments count towards your plan's file-upload limits.
Audio uploads must be audio. MP4 and WebM files that are identified as video are not accepted as audio uploads.
For data analysis, ChatGPT writes and runs Python code in a notebook environment that can use the files you give it, and that environment cannot make external web requests.

So you can ask it about a video and sometimes get a useful answer. You cannot rely on it to have seen every second or heard every word, which is exactly what an edit needs.

What is ChatGPT good for in video editing?

The words around the edit, where it is genuinely quick:

Scripts and outlines, including tightening a rambling draft before you record.
Titles, descriptions and tags, from a transcript you paste in.
Chapters, from a transcript with timestamps.
A paper edit: give it a transcript and ask which lines to keep for a three-minute version. You still cut the video yourself.
Commands for tools you run, such as a command line to convert or trim a file. Check them before you run them.

What these have in common is that you do the cutting. A chat can tell you what to do; it cannot see your timeline, play your footage back, hear a stumble or render the result.

What is the difference between a chat and an agent that edits?

Tools. A chat assistant answers in text. An agent is the same kind of assistant given tools it can call, so it can act: open a recording, read it, cut it, render it, check the result and try again.

Job	A chat on its own	An assistant driving a real editor
See every word and pause	From a pasted transcript, if you give one	Reads the recording itself, on your Mac
Choose between takes	Suggests, from text	Compares every take and keeps the best
Cut the video	Gives you instructions	Makes the cuts, with a reason for each
Captions	Writes caption text	Burns in captions timed to the words
Render and check	Not possible	Renders, then checks the render
Your footage	Uploaded with the chat	Stays on your Mac

A young woman holding a cup of tea watches an open laptop at her kitchen table, one hand at her chin. — You describe the edit in plain words; the editor does the cutting on your own Mac and reports back.Photo by Helena Lopes on Pexels

How does an AI assistant drive a real editor?

Through MCP, the Model Context Protocol: an open standard that lets an assistant use another program's tools. The editor publishes what it can do; the assistant calls those tools in the order a job needs.

Sharpy comes with an MCP server, sharpy-mcp, that offers its whole editor:

105 tools, including the five that carry an edit from start to finish: read_format, set_format, read_recording, plan_beats and assemble.
The working material as resources: the media, the transcript, the timeline and the edit decisions.
16 workflow prompts, ready-made plans such as editing a recording to match a reference.
Progress as it works, so a long render reports back instead of going quiet.

You ask in plain language, "cut today's recording to a 60-second vertical short, and keep the second take of the intro", and the agent starts with the workflow tool, which sets out the steps for the job, then reads, plans, cuts and renders. You and the agent edit the same project: its steps join the window's undo history, and it is told what you changed. Editing with Claude or Codex through MCP has the setup, one command in Claude Code or one entry in a settings file.

Can I use ChatGPT itself with Sharpy?

Sharpy's server runs on your Mac and talks to the assistant through the standard input and output of a local process; it opens no network port. Clients that start local servers connect to it directly. Codex, OpenAI's coding agent, is one: its documentation lists local STDIO servers alongside Streamable HTTP ones, set in ~/.codex/config.toml. Claude Code and Claude Desktop are others.

ChatGPT connects to MCP servers differently. OpenAI's developer documentation describes adding a custom MCP server to ChatGPT through a public HTTPS endpoint, or through OpenAI's Secure MCP Tunnel, which can front a local server without exposing it to the internet. That is a developer's setup rather than a settings switch, and Sharpy's own guide covers Claude and Codex. If you want an OpenAI assistant in charge of the edit, Codex is the direct route.

What stays on your Mac?

Your footage, audio, transcripts and projects. Sharpy reads and renders everything on your Mac, so there is no video upload and no per-minute charge: one plan, unlimited edits, however you drive it.

The conversation is another matter. Whatever you type, and whatever Sharpy's tools report back to the assistant, such as parts of the transcript and the edit decisions, goes through the assistant's provider under its terms. If a project must not leave your Mac at all, edit in Sharpy's window directly. Editing with AI without uploading your footage explains the difference.

Frequently asked questions

Can I use ChatGPT to edit a video?

You can use it to plan an edit: a script, a paper edit from a transcript, titles and chapters. The cutting needs an editor, either one you operate or one an assistant drives through MCP, such as Sharpy.

Can ChatGPT watch a video?

Partly. OpenAI says ChatGPT can analyse uploaded videos with tools, but that it may not analyse the entire video or accurately interpret its audio.

Can ChatGPT add captions to a video?

It can write caption text or a subtitle file from a transcript, but it does not burn captions into your video. An editor times and draws them; Sharpy does both from the edit's own words.

Is there an AI that edits videos for you?

Yes. An agentic editor such as Sharpy edits the whole piece from raw footage and gives a reason for every cut. What an agentic video editor is explains how.

Does ChatGPT make videos?

Making new footage from a prompt is a different job, done by video generators. AI video editors and generators explains the difference.

Sources

Checked on 18 October 2026.