Can ChatGPT Make Videos? We Tested 3 Ways After Sora
Yes, but not with Sora. OpenAI shut down the Sora app on April 26, 2026 and the video API on September 24, 2026. ChatGPT can still hand you an MP4 by writing code that draws every frame, on the web and in the desktop app. In our test those videos used a single diagram, cut text in without animation and stayed still most of the time. For animated motion graphics, connect a video app to ChatGPT, such as iArt.
iArt is our product. On November 17, 2026 we gave ChatGPT one request three ways and kept every result as it came back. The videos, times and costs below are from that test.
What happened to Sora
Sora was OpenAI's video model, and for a while it was how people made videos in ChatGPT. OpenAI's Sora 2 page now opens with the line "As of April 26, 2026, the Sora product is no longer available." The developer side followed: OpenAI's deprecations table lists the Videos API and the sora-2 and sora-2-pro models as shut down on September 24, 2026, with "no one-to-one replacement".
So ChatGPT no longer has a video model of its own. What it still has is the ability to write and run code, and apps you can connect to it. Those are the routes we tested.
The test: one request, three routes
We picked a request that needs more than one chart: a step-by-step explainer with a diagram, a signal that travels, a result that has to appear at the right moment, and text that must be readable. Every route got exactly these words:
Make a 20-second 16:9 explainer video for YouTube: "What happens in the 2 seconds after you tap your card." Show the path step by step with a simple diagram: card → payment terminal → card network → your bank checks the balance → "Approved" back on the terminal. Label each step, animate the signal moving along the path, and hold each step long enough to read. End on the line "All of this in about 2 seconds."
The three routes:
- ChatGPT on the web (chatgpt.com, free account), no app connected.
- The ChatGPT desktop app with full access to the computer, no app connected.
- ChatGPT with iArt connected as a custom MCP server, desktop app.
Route 1: ChatGPT on the web
ChatGPT answered with "Video created" and a download button. It had written a Python program in its sandbox that draws each frame of the diagram and joins the frames into an MP4: 20 seconds, 1920 × 1080, 30 frames per second, no sound.
Against the request: the five steps, the moving signal and the closing line are there. "Approved" is lit from the first second instead of arriving at the end, two labels run into each other ("PAYMENT TERMINALCARD NETWORK"), and the captions are too small to read on a phone. No text animates: captions change with a hard cut.
Route 2: the ChatGPT desktop app
With full access to the computer, ChatGPT worked for 2 minutes 10 seconds. It wrote a Python script on the machine, rendered the frames, encoded the video and checked it, then returned a 20-second, 1280 × 720 MP4 at 20 frames per second, without sound.
It adds a title card and a clean layout, and the signal follows the path, including the way back. It still misses the "Approved" moment (the last step only says "Decision returns"), the small labels are hard to read on a phone, and every piece of text cuts in without animation.
Route 3: ChatGPT with iArt
With iArt connected, ChatGPT sent the request to iArt as a brief, checked on it while iArt wrote and rendered the animation, and came back with an MP4, a share link and a link to edit the project: 20 seconds, 1920 × 1080, 30 frames per second. It used $0.77 of iArt credit.
Every part of the request is on screen: each step is labeled, the nodes light up one by one as the signal reaches them, the bank runs a check, and the terminal screen turns green with "Approved" before the closing line. The title builds word by word and the step captions fade in and out at a size you can read on a phone.
The iArt video is the one at the top of this article.
Side by side
We checked each video against the request, line by line, and measured how much of the picture moves.
| In the request | ChatGPT web | ChatGPT desktop | ChatGPT + iArt |
|---|---|---|---|
| Card → terminal → network → bank, each step labeled | Yes (two labels overlap) | Yes | Yes |
| The signal moves along the path | Yes | Yes | Yes |
| "Approved" back on the terminal | No (lit from the start) | No | Yes |
| Each step readable (on a phone) | No, captions too small | No, labels too small | Yes |
| Ends on "All of this in about 2 seconds." | Yes | Yes | Yes |
| Text animation | None (hard cuts) | None (hard cuts) | Word-by-word title, fading captions |
| Frames where the picture changes | 5% | 2% | 29% |
| Video | 1080p, 30 fps | 720p, 20 fps | 1080p, 30 fps |
| Cost | Free account | Included in the ChatGPT plan | $0.77 of iArt credit |
How we judged the motion
"Frames where the picture changes" compares every frame with the one before it. In both ChatGPT videos, a small dot and a highlight are the only things that move; in the iArt video, text, nodes and signals animate.
We also gave the three videos to Gemini 3.8 Flash and asked, without saying which was which, which had the better motion design, showing each pair in both orders. The iArt video won all four of its comparisons, and the desktop video beat the web one both times. For the desktop video, Gemini noted that "every text element appears via instantaneous hard cuts".
Which route should you use?
- A quick chart of one number, for free: ask ChatGPT directly. It can render a simple animated chart, as long as you check the numbers and do not need moving text.
- Motion graphics: explainers, kinetic typography, logo reveals, data stories and series. Connect a motion graphics app such as iArt, which animates the text and the elements and gives you an MP4 and a share link.
- Generated footage of people and places: the ChatGPT plugin directory lists video apps such as VideoGen, VEED and invideo; each works with its own account and pricing.
How to add iArt to ChatGPT
In ChatGPT, open Customize → Plugins → Add → Add custom MCP server, name it iArt, paste https://app.iart.ai/api/mcp, set Authentication to OAuth and sign in to iArt. Then ask for a video in the chat. The step-by-step guide with screenshots walks through each screen; making videos needs an iArt Pro, Ultra or Max plan.
Using Claude instead? The same connector works there: see the iArt video MCP server.
FAQ
Can ChatGPT make videos?
Not with its own video model any more: OpenAI discontinued the Sora app on April 26, 2026 and the video API on September 24, 2026. ChatGPT can still write code that renders a simple MP4, and it can make videos through apps you connect, such as iArt.
Can ChatGPT make a video for free?
On a free chatgpt.com account, ChatGPT wrote code that rendered a 20-second 1080p MP4 for our request. The result was a static diagram with a moving dot and no text animation.
Is Sora still available in ChatGPT?
No. OpenAI says the Sora product has been unavailable since April 26, 2026, and its developer docs list the Videos API and the sora-2 models as shut down on September 24, 2026.
Can ChatGPT animate text?
Not in our test: in the videos ChatGPT rendered with code on the web and in the desktop app, every piece of text appeared with a hard cut. With iArt connected, the title built word by word and the captions faded in and out.
How do I make an animated explainer in ChatGPT?
Connect a motion graphics app. With iArt, add it as a custom MCP server in ChatGPT, sign in, and describe the video in the chat; you get an MP4, a share link and an editable project a few minutes later.
How much does a video cost with iArt in ChatGPT?
It uses iArt credit at the same prices as in the iArt app. Our 20-second explainer cost $0.77. Making videos needs a Pro, Ultra or Max plan.