Can ChatGPT Edit Videos? What ChatGPT, Claude & Gemini Can Do (2026)

Not on its own. As of October 2026, ChatGPT, Claude and Gemini can plan an edit, write scripts, shot lists and FFmpeg commands, and read still frames you export; Gemini can also analyze uploaded video and generate or alter short clips. None of them cuts your project on a frame-accurate timeline unless you connect an editor through MCP or use an agentic video editor.

This is the chat-assistant side of the broader question can AI edit videos?. Below: what each assistant can do according to its maker's documentation, what none of them can do, how MCP connectors change that, and how agentic editors differ.

Facts checked October 20, 2026 against Anthropic's documentation and help center, OpenAI's developer documentation, and Google's Gemini help and product pages (links under Sources). OpenAI's ChatGPT help center couldn't be checked for this update, so the ChatGPT entries rely on OpenAI's developer docs; check the app for consumer features, which change often. Full disclosure: Loopdesk is our product.

What can ChatGPT, Claude and Gemini do with video?

All three can write and reason about an edit, read images and help with code. Gemini is the only one of the three whose app documents uploading and analyzing video, and it also generates short clips. None documents editing your own project on a timeline without an external tool.

"Write scripts" and "write FFmpeg commands" are general text abilities, so check every command before you run it on footage you care about.

Can ChatGPT edit videos?

Not directly. OpenAI's developer documentation says "All latest OpenAI models support text and image input, text output," and its vision guide covers images only. ChatGPT can plan an edit and write commands, but OpenAI doesn't document it cutting your video files.

What OpenAI does document:

• Code, not video. The API's Code Interpreter "allows models to write and run Python code in a sandboxed environment." Its supported-file table lists formats such as PDF, CSV, XLSX, DOCX, PPTX, JSON, images and ZIP, and no video formats.
• Sora's API is gone. OpenAI's developer docs say "The Sora 2 models and Videos API were shut down on September 24, 2026 and are no longer available," with "no one-to-one replacement API." We couldn't confirm whether video generation remains in the consumer Sora or ChatGPT apps.
• MCP connections. OpenAI's Apps SDK docs describe adding a server from ChatGPT's Plugins page with "Add custom MCP server", after a risk warning you must accept. That is how ChatGPT could reach a video tool, if someone has built a connector for it.

What can Claude do with video?

Claude can't watch video. Anthropic's vision docs list JPEG, PNG, GIF and WebP images, say animations are unsupported and only the first frame is used, and state that "Claude is an image understanding model only": it can't generate or edit images, and Anthropic documents no video generation.

Claude is still useful around an edit:

• Frames and transcripts. Export a frame per scene and the transcript, and Claude can describe shots, spot continuity problems and propose an edit decision list. claude.ai accepts up to 20 images per message.
• Code and files. On every plan, Claude can run code in a sandboxed container to create and edit files (owners control this on Team and Enterprise), with a 30 MB limit per file. Anthropic lists documents, spreadsheets, slides, PDFs and data files, not video.
• Custom connectors. Claude supports custom connectors using remote MCP on Free (one connector), Pro, Max, Team and Enterprise plans, across claude.ai, Claude Desktop and the mobile apps. Anthropic warns that these connect to services it hasn't verified, that "malicious MCP servers may include hidden instructions," and that connectors can read, change or delete data in connected apps.

What can Gemini do with video?

Gemini documents the most video features of the three. The Gemini app analyzes uploaded videos (each up to 2 GB, 5 minutes total, or 1 hour with Google AI Pro or Ultra) and, with a Google AI subscription, generates 10-second clips with native audio. It can also make generative edits to clips, but it isn't a timeline editor.

• Generation and generative edits. Google says the Gemini app now uses Gemini Omni 1.1 Flash, which "will now replace the previous Google Gemini Veo 3.1 model." It turns text, photos (up to five) or video into video, and Google's page says: "Swap characters, adjust lighting, stabilize the video, or modify the background with a chat." Generated videos carry SynthID, Google's "imperceptible watermark".
• The API for developers. Gemini's API documentation says models with a 1M-token context window can process videos up to 3 hours long (1 hour at high media resolution), sample visual content at 1 frame per second, and accept timestamps in MM:SS format.
• MCP. Gemini CLI can connect to MCP servers and asks for confirmation before running their tools unless you mark a server as trusted.

The 1-frame-per-second sampling is worth remembering: a model that samples once a second can describe a scene well and still miss a cut point by a fraction of a second. Check timestamps yourself before you cut.

What can't chat assistants do with your video projects?

They can't open your project, trim clips frame-accurately, mix audio or render a finished export on their own. Even Gemini's video features generate new clips rather than editing your timeline. Story, pacing and brand judgment stay with you.

• No timeline access. Without a connector, an assistant only sees what you upload or paste. It can't move a clip in your Premiere, DaVinci Resolve or Loopdesk project.
• No guaranteed precision. Timestamps from a chat answer are suggestions; verify them on the timeline.
• No final render. None of the three documents rendering your edited video file.
• Your data, their terms. Uploading client or confidential footage to any assistant is governed by that service's data terms, so read them first.

How can a chat assistant drive a real video editor?

Through an MCP connector. MCP lets an assistant discover and call tools exposed by another program, but as our guide to MCP for video editing puts it, MCP "does not edit video by itself." The connector and the editor behind it decide what's possible: reading a project, staging a change, trimming, rendering or publishing.

Community MCP servers exist for DaVinci Resolve and Premiere Pro, maintained by individuals; our AI video editing agents guide lists them and their requirements, such as Resolve Studio for external scripting. Claude's custom connectors, ChatGPT's custom MCP servers and Gemini CLI can all act as the client.

If you try one, work on a duplicate project, start with read-only tools, approve every change that alters the timeline, and check the exported file rather than trusting the assistant's summary. If you'd rather automate edits in code, compare video editing APIs instead.

How is an agentic video editor different from a chatbot?

An agentic video editor has the timeline, media engine and renderer built in, so its AI can plan a multi-step edit, apply it to your footage and leave a result you can review and undo. A chat assistant on its own advises on your footage rather than editing it. See what agentic video editing is for the full definition.

The input is the same either way, plain English, which is what natural-language editing means. The difference is what happens next. In Loopdesk, the AI agent Aura has Quick Actions (Analyze Video, Generate Highlights, Suggest Edits and Transcribe) and also takes free-form prompts. Suggest Edits produces an edit plan you review and apply. Aura handles silence and filler-word removal, rough cuts, multicam switching by speaker, highlight clips, reframing to vertical, captions in 108 languages and batch editing across 50+ videos, all in the browser. Loopdesk has no free plan or trial, and story decisions still need your review. For one-click automation without an agent, see what auto edit is.

How do you use ChatGPT, Claude or Gemini in a video workflow today?

Use a chat assistant for the thinking and text around an edit, and an editor for the cutting. These five jobs suit a chat assistant:

1. Script and shot list. Give the audience, length and key message, and ask for a script with a shot list and B-roll ideas.
2. Edit decision list from a transcript. Paste a timestamped transcript and ask which segments to keep for a 3-minute version, with in and out points. Apply them in your editor.
3. FFmpeg commands. Ask for a command, then check it. A typical trim looks like this:

ffmpeg -i input.mp4 -ss 00:01:30 -to 00:02:15 -c:v libx264 -c:a aac clip.mp4

Re-encoding as above gives an accurate cut; copying streams without re-encoding is faster but cuts on keyframes, so the start can shift.

4. Titles, descriptions and chapters. Paste the transcript and ask for a title, description and chapter list.
5. Footage review. In the Gemini app, upload a short clip and ask where topics change or which moments to clip, then verify each timestamp.

Which is the best LLM for editing videos?

None of them edits a timeline by itself. For understanding footage, Gemini is the only one of the three whose app documents video upload and analysis; for scripts, edit plans and code, all three work. For finished edits, use an agentic video editor or connect an editor through MCP and review every change.

Sources

Read October 20, 2026.

• Anthropic, Vision: https://platform.claude.com/docs/en/build-with-claude/vision
• Claude Help Center, supported uploads: https://support.claude.com/en/articles/8241126-what-kinds-of-documents-can-i-upload-to-claude-ai · creating files: https://support.claude.com/en/articles/12111783-create-and-edit-files-with-claude · custom connectors: https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp
• OpenAI, models: https://developers.openai.com/api/docs/models · images and vision: https://developers.openai.com/api/docs/guides/images-vision · Code Interpreter: https://developers.openai.com/api/docs/guides/tools-code-interpreter · video generation: https://developers.openai.com/api/docs/guides/video-generation · deprecations: https://developers.openai.com/api/docs/deprecations · connecting an MCP server to ChatGPT: https://developers.openai.com/apps-sdk/deploy/connect-chatgpt
• Google, Gemini Apps uploads: https://support.google.com/gemini/answer/14903178 · video generation: https://gemini.google/overview/video-generation/ · Gemini API video understanding: https://ai.google.dev/gemini-api/docs/video-understanding · Gemini CLI MCP servers: https://geminicli.com/docs/tools/mcp-server/

Frequently Asked Questions

Can ChatGPT edit videos?

Not directly. OpenAI's documentation describes text and image input, not video editing. ChatGPT can write scripts, shot lists and FFmpeg commands, and with a custom MCP server it can call tools in an editor that exposes them. For actual cuts, use a video editor.

Can Claude watch or edit videos?

No. Claude accepts images (JPEG, PNG, GIF and WebP), not video files, and Anthropic says it can't generate images. It can analyze frames you export, write edit plans and code, and drive an editor through a custom MCP connector.

Can Gemini edit videos?

Partly. The Gemini app analyzes uploaded videos (up to 5 minutes total, or 1 hour on Google AI Pro or Ultra) and generates 10-second clips with audio, including generative "video to video" edits. It doesn't edit your project on a timeline.

What is the best LLM for editing videos?

None edits a timeline by itself. Gemini is the only one of the three whose app documents video analysis, and all three help with scripts, shot lists and FFmpeg commands. For finished edits, use an agentic editor or connect an editor through MCP.

Can ChatGPT make videos with Sora?

OpenAI's developer docs say the Sora 2 models and Videos API were shut down on September 24, 2026, with no one-to-one replacement API. We couldn't confirm the status of video generation in the ChatGPT or Sora apps, so check the app.

How can an AI assistant control Premiere Pro or DaVinci Resolve?

Through an MCP connector. Community servers exist for both, and Claude, ChatGPT and Gemini CLI can add MCP servers. The connector defines what's possible, so work on a duplicate project and approve every change.

---

Want an AI that edits the timeline instead of describing it? Try Loopdesk - ask Aura to cut silences, caption and make vertical clips, then review every change before export.