If you've gone looking for video in ChatGPT recently and found it missing, you haven't got the wrong menu. OpenAI closed the Sora app and website on 26 April 2026 and switched off the Sora API on 24 September. Sora used to come with the Plus and Pro plans. It doesn't any more, and right now OpenAI doesn't offer a video product to consumers at all.

That sounds like the end of video in ChatGPT. It isn't, because Sora was never the part that edited your footage. Sora made new video from a prompt. Editing the video you've already shot was always going to need a proper editing tool, and that's the part that has quietly got better.

01 / The Adobe plugin

Think: an editor you brief in one sentence.

In August, Adobe brought more than 70 of its tools into ChatGPT as a single plugin, drawing on Premiere, Photoshop, Express, Firefly, Lightroom and the rest. For video, it does three useful things. It can make a quick cut, pulling the strongest moments out of a longer video into a short edit. It can turn those moments into a highlight reel. And it can resize a video for different platforms, so one horizontal clip becomes versions for Reels, Shorts and TikTok.

To turn it on, open Plugins in the ChatGPT sidebar, search for Adobe and connect it. After that you can call it from any chat by typing @Adobe. You don't need an Adobe account to try it, because it runs in a guest mode first, but signing in gives you the more detailed editing. Your results save to Creative Cloud, and if you want to keep refining an edit you can take it into the full Adobe apps and carry on from there.

02 / The one-job route

Think: you have an editor already. You just need one thing done.

Plenty of people don't want an AI to make the edit. They edit in CapCut or Premiere and they're happy with it. What they want is help with the one job their editor is bad at. That's where Higgsfield comes in. It's known mostly for generating AI video, but it can also upscale footage, remove the background from a video and reframe a clip to a new shape by extending the picture rather than cropping into it. It connects to ChatGPT from the Plugins directory. It needs a paid Higgsfield subscription, and each job uses credits from your Higgsfield account.

03 / What to ask for

Think: how long, which platform, what stays, what goes.

The clearer the brief, the less you redo. A few that work:

• "@Adobe Cut this ten-minute walkthrough down to 30 seconds for Reels, and keep the final reveal."

• "@Adobe Make vertical versions of this video for Reels, Shorts and TikTok."

• "Upscale this clip so it holds up on a bigger screen."

• "Reframe this horizontal shot to vertical without cropping off the sides."

04 / Where ChatGPT is still the best part

Think: everything either side of the edit.

The editing tools do the cutting. ChatGPT is still the better collaborator for everything around them. Before you shoot, it can turn a vague idea into a shot list so you come home with what you need. After, it can write the hook for the first two seconds, the caption and the title. If you paste in a transcript, it can tell you which line should open the video and which tangent should go. None of that is new, and it's still where most of the value is.

Where it still needs you

Picking the moments is still your job, particularly for visual content. AI editing tools lean on what's said and on a sample of frames, so they tend to favour talking over visuals. If your video is a match, a sunset or b-roll with no dialogue, tell it the timestamps you want.

Check every export before you post it. Make sure it plays, the reframe hasn't cut off a face or a product, and the captions say what was actually said.

And no tool can fix the footage itself. Shaky, dark, badly framed video comes back as shorter shaky, dark, badly framed video. The edit is only as good as what you shot. Where ChatGPT does get better is once it knows you, and that's what AI for Creators is for: setting ChatGPT or Claude up so it already knows your work, your style and how you like to work, before you ask it for anything.

And in Claude

Claude works the same way, with the same Adobe tools through its own Adobe for creativity connector, plus Higgsfield and Descript, which is built for talking-led videos you edit by editing the transcript. If you're choosing between the two, the honest answer is to use whichever you already pay for. The full Claude version is here: How to Edit Video With Claude.

This reflects the tools as they exist in October 2026. Plugins change and names move, but the principle holds: let the AI handle the instructions and the repetitive jobs, and keep the judgement for yourself.