Yes, ChatGPT can make animations in two ways: short looping GIFs it creates itself, a feature OpenAI says is still rolling out, and animation code (CSS, SVG, Lottie or Manim) that you preview or render yourself. What it doesn’t give you is a graphic you can keep editing inside a video project, and Sora, OpenAI’s video app, has been discontinued. Below are seven routes, what each one returns, and how to make an animated graphic you can still revise in your video.
Can ChatGPT make animations?
Yes, but the route depends on what you need back. Ask for a GIF when you want a loop. Ask for code when you need exact control over text, shapes or timing. For a graphic inside a video that you’ll revise later, use an editor that keeps the graphic in its project.
As of November 12, 2026, OpenAI documents GIF creation as a rollout, so not every account has it yet. ChatGPT Work can also run code and shell commands in the environments your account can access, which makes code-based animation possible but not guaranteed in every chat.
| What you get | Preview it in | Keep for revisions |
|---|---|---|
| Native GIF | Any GIF viewer | Prompt and starting image |
| HTML/CSS animation | A browser | HTML and CSS file |
| Animated SVG | A browser | SVG file |
| Lottie JSON | A Lottie player | Validated JSON |
| Manim scene | Rendered MP4 | Python scene and assets |
| Generated clip | Its own tool | Prompt and input images |
| ChatCut graphic | ChatCut timeline | The graphic in its project |
A GIF or video file is a finished render. It doesn’t come with layers, character rigs or the source used to make it.
Seven ways to make animations with ChatGPT
We wrote these prompts for this guide and haven’t run them, so open and test whatever comes back.
1. A native looping GIF
Use it for: a small repeating illustration, reaction or icon. OpenAI’s GIF instructions say to describe the loop, or upload a still image and ask ChatGPT to animate it.
What you get and what to check: a GIF file. Watch it loop several times for a jump at the seam, unwanted camera movement or a subject that changes shape. ChatGPT can’t edit an animated GIF you upload in this workflow; start from a still image instead.
2. A CSS title animation
Use it for: a title, loader or short text sequence whose timing you want to tweak. CSS animations move between style states using keyframes, so ask for one complete file and open it in a browser.
What you get and what to check: an HTML file. Open it, replay it, and change the title once to see whether the layout survives. A browser preview isn’t a video file; if you need one, plan a separate screen-recording or render step.
3. An animated SVG diagram
Use it for: a labeled process, a moving arrow or a simple shape diagram. SVG keeps each element in the file, and SMIL animation can animate their numeric attributes. Name the sequence and the labels instead of asking for “an animated explainer”.
What you get and what to check: an SVG file. Preview it in the browser you’ll use, and check arrow direction, label fit and the final frame.
4. A Lottie JSON animation
Use it for: a small app or website animation when the destination expects Lottie. Lottie is a JSON-based animation format, so asking ChatGPT for one is a code-writing task, not a built-in Lottie export.
What you get and what to check: a JSON file. Validate it, then load it in the player your app uses. Valid JSON can still play incorrectly, so check timing and looping there.
5. A Manim scene rendered to MP4
Use it for: math, diagrams or shape sequences you want to describe in Python and render as video. The Manim quickstart shows scene code rendering to an MP4. You need Manim installed somewhere; the Python in the chat is only the source.
What you get and what to check: a Python file, then an MP4 once you render it. Fix any errors, then watch the actual video for the shape change, the duration and the final hold.
6. Artwork and a scene plan
Use it for: preparing an animation before you pick the tool that moves it. ChatGPT can generate and edit still images and write a plan of what happens in each shot.
What you get and what to check: a plan and still images, not a moving file. Check that characters, colors and style match from scene to scene before you animate any of them.
7. A clip from an animation generator
Use it for: movement you want a video model to create, starting from ChatGPT’s artwork or plan. The clip itself is made in another tool; Adobe Firefly, for example, turns images and prompts into animated clips.
What you get and what to check: a rendered clip. Look for details that change between frames and text that turns unreadable. Unless the tool says otherwise, treat the clip as finished media, not an editable project.
A note on Sora tutorials
OpenAI says the Sora web and app experiences ended on April 26, 2026, and the Sora API on September 24, 2026. If a tutorial tells you to open Sora, swap that step for one of the routes above. ChatGPT’s GIF feature doesn’t bring Sora back.
Keep animation editable in your video
Decide what you’ll need to change later. To tweak a CSS keyframe, keep the CSS. To change a chart label while reviewing a cut, you need the graphic to live in the video project. Dropping a rendered clip into an edit isn’t the same as being able to change what’s inside it.
In ChatCut, Motion Graphics cover titles, lower thirds, callouts, data charts, diagrams and animated overlays, and they stay editable on the timeline. Our AI motion graphics guide covers that workflow in depth.
Prepare the content and placement
Have the final words and numbers ready, plus where the graphic goes on the timeline, how long it lasts and where it sits in the frame. Add colors or a style reference if they matter. You can make the request in ChatCut’s AI panel or from a desktop agent connected as described below. From an external agent, the project needs to be selected and ChatCut connected, but the project window doesn’t need to be open.
Inspect and revise one property
Play the chart with the video around it. Check that it starts at 00:12, shows the right values and stays readable while the bars move. Then ask for the value and duration change and compare.
In ChatCut Desktop 0.3.13 or later, you can edit elements directly: select one Motion Graphic on the timeline, then select an element to change its text, font or color, or drag it to reposition it. For structural or animation changes without a direct control, or on other surfaces, ask the Agent. ChatCut doesn’t document importing your own CSS, SVG, Lottie or Manim files as editable Motion Graphics, so keep code-based animations as separate files.
Choose the final delivery
Export the whole video when the graphic belongs in the finished edit. For separate transparent overlays, Export → Graphics renders each motion graphic on the active timeline as its own ProRes 4444 .mov with an alpha channel, which requires Pro. Keep the project too, so you can reopen it and change a value later.
On cost: manual timeline edits and uploads don’t use ChatCut credits. Motion-graphics generation in ChatCut’s built-in Agent uses credits based on model usage. From Claude Code or Codex, the agent’s reasoning is covered by your Claude or ChatGPT plan, and ChatCut credits apply when the workflow calls ChatCut-hosted generation; check any confirmation card that appears. See the credits policy.
Connect Codex to ChatCut Desktop
Follow the agent installation guide. Start a new conversation in Codex on your computer, then paste and send:
Sign in to ChatCut if it asks. Then start another new conversation: agents load ChatCut’s tools when a conversation starts, so the installation conversation doesn’t get them. In the new conversation, ask the agent to use ChatCut and name the project and graphic you want. ChatCut can’t be installed by pasting this into the ChatGPT or Claude websites. The ChatGPT plugin page shows the setup screens.
Can Claude make animations too?
Yes. Anthropic documents Claude Motion as a beta on Team and Enterprise plans, with MP4 export. It’s a separate tool from an agent connected to ChatCut: Claude Motion makes its own animations, while a connected agent works on a ChatCut project. Access to one doesn’t give you the other.
Check the animation before delivery
Whichever route you use, check these before you call it done:
- File type. Open the actual GIF, source file or video. A description of an animation isn’t the deliverable.
- Readability. View labels at the size people will see them, and hold important text long enough to read aloud.
- Timing. Check the start, the end, the loop seam and any moment that must line up with narration.
- Accuracy. Compare every name, number, arrow and chart value with your content.
- Dependencies. Reopen the source with its fonts and assets available and fix any errors.
- Revisions. Change one word or value, preview again, and check what else moved.
Make one graphic you can revise
Start with a title or chart in a real video. Give the final content, timing and placement, check the first result, then change one value before you export. That one revision tells you whether you have the editing control your next project needs.
