Can ChatGPT make videos? Until spring 2026 the answer was "yes, with Sora". That answer is out of date. Today ChatGPT can plan a video, write its script and make still images, and it cannot hand you generated footage. There is one route to a real video file that runs on a ChatGPT plan, and it goes through code.
Everything below was checked on OpenAI's own pages on November 19, 2026.
What do OpenAI's pages say today?
| Question | What OpenAI's pages say | Where |
|---|---|---|
| Is Sora still in ChatGPT? | The Sora web and app experiences were discontinued on April 26, 2026 | Help article "What to know about the Sora discontinuation" |
| Can developers still call it? | The Videos API and the sora-2 and sora-2-pro models shut down on September 24, 2026. Replacement: "No one-to-one replacement" | API deprecations page |
| What do the plans include now? | Image creation and Codex on every tier, in growing amounts from Free to Pro. Video generation is not listed | ChatGPT pricing page |
| What about videos made before? | OpenAI pointed users to an export page and said Sora data would be deleted after the shutdown | The same help article |
So the old guides that explain Sora limits by plan (so many videos a month on Plus, longer clips on Pro) describe a product that no longer exists. If you read one, check its date.
What can ChatGPT still do for a video?
Plenty of the thinking, none of the rendering.
- The script. Hook, beats, ending, at the length you ask for.
- The storyboard. A shot list with what is on screen for each line.
- Still images. ChatGPT's plans include image creation, so it can make frames, backgrounds or a thumbnail.
- Prompts for another tool. It will write the prompt you paste into a video generator. AI video prompts shows what a good one contains.
All of that still leaves you with text and pictures. Someone, or something, has to make the file.
Can AI make videos at all, then?
Yes, in three different ways, and they give three different kinds of video.
| Route | What it makes | Examples | What it costs you |
|---|---|---|---|
| Footage models | Clips that look filmed, a few seconds each | Veo, Kling and others, each on its own site | Credits per second, and less control over text and facts |
| Avatar tools | A presenter reading your script | Covered in HeyGen alternatives | A subscription, usually by the minute |
| Code-to-video | Animation and motion graphics drawn by a program | Codex or Claude Code plus a renderer | Your AI plan, and a renderer on your computer |
For footage, Sora alternatives compares what is left now that Sora is gone. The third route is the one a ChatGPT subscriber already has.
How does prompt to video work with Codex?
Codex is OpenAI's coding agent. Its documentation lists where it runs: the ChatGPT desktop app, the web, a command line tool, an IDE extension and the cloud. The pricing page lists it on every ChatGPT plan, limited on Free and expanded on Plus and Pro.
A coding agent can do what a chat cannot: write files and run programs. That is enough to make a video.
- You describe the video.
- Codex writes it as a web page: HTML for the layout, JavaScript for the motion, every movement tied to a time.
- A renderer opens the page in a hidden browser and saves one picture per frame.
- An encoder joins the frames and the sound into an MP4.
What you get is exact. Text is real text and never comes out garbled. A number on screen is the number in the code. What you do not get is filmed-looking footage: this is drawn animation.
You can set this up yourself with open-source tools if you are comfortable in a terminal. What is HyperFrames explains the rendering side.
What does Easy Reels add on top of Codex?
Easy Reels is a desktop app for Mac and Windows that does the setup for you. Under the hood each video is an HTML page rendered to MP4. The part that writes and directs runs on the AI you pick in Settings, and Codex on your ChatGPT plan is one of the choices, with its own model and effort picker.
You pick one of five ways to make a reel:
- Pixel art: a 3D scene drawn in pixels that cuts a thing open, slows it down and names its parts.
- ASCII: the same kind of scene drawn in text characters, up to 4K.
- Dots: two colours you choose and patterns of dots, like an old newspaper print.
- Motion: a narrated reel from nothing, with script, voice, animated scenes, captions, music and sound effects.
- Your videos, edited: it cuts pauses, stumbles and repeated takes, frames vertical and adds captions. Can ChatGPT edit videos? covers that side.
Then you type an idea, press Improve my idea to turn a few words into a full request, and press Generate. With Codex the app looks facts up on the web before it writes. The video is made on your computer, vertical 9:16 by default. A voice is included and runs locally.
It costs $39 once, with a 7-day refund and no credits: pricing. There is more on the approach at AI video generator.
When is this the wrong route?
- You want realistic footage. Easy Reels is not a diffusion model. It will not make a drone shot of a beach or a person walking down a street.
- You want an avatar or lip sync. It has neither.
- You want to work on a phone or in a browser. It is a desktop app, with no cloud and no mobile version.
- You want it to post for you. It makes the file. Scheduling and posting are yours.
If none of those apply, and the video is an explanation of how something works, code-to-video is the route where a ChatGPT plan is enough.
Questions
Can ChatGPT make videos for free?
Not footage, on any plan, as of November 19, 2026. The free plan lists limited Codex access, so a small code-to-video test is possible, within that limit.
Is Sora coming back to ChatGPT?
OpenAI's deprecations page lists "No one-to-one replacement" for the Sora models. We found no announced date on OpenAI's pages on the day we checked.
Can ChatGPT make a video from images?
It can make the images. Turning them into a video needs another program: an editor, a footage model that animates stills, or code that moves them.
Does Easy Reels need a ChatGPT plan?
No. It runs on Codex, on Claude Code, or on a free local model downloaded in the app. The local model works offline and makes Motion reels from the scene library and edits of your videos.