How to Make Motion Graphics in After Effects With ChatGPT

You might be tired of AI motion graphics that arrive as one flat video you cannot touch. The title is spelled wrong, the number should say 84 not 48, and the only fix is to generate the whole thing again. There is now a way round it. ChatGPT can open After Effects, build the animation inside a real project and hand it back with every layer editable.

This is how that works, what it is good at, what a piece costs in plan usage, and where you still finish by hand. All of it is as of October 2026.

You get a project file, not a flat video

A Seedance clip with animated graphics in it looks fine until the client changes a word. After Effects work is built from layers, keyframes and compositions, so changing a word means clicking a text layer and typing.

When ChatGPT Astra 6 drives After Effects, it makes those layers. It names and organises them, and it puts the animation on keyframes you can open and move. We have changed a name on a stat card, made a text layer slide in more slowly and altered a colour, all without regenerating anything. If you cannot find where something lives, take a screenshot and ask Astra. It built the project, so it can tell you which layer to open.

Setting it up takes about ten minutes

1. Install the ChatGPT desktop app. The browser version cannot reach the programs on your computer.
2. Switch the app to its work mode and choose Astra as the model.
3. Turn on the computer use plugin, then give it the screen recording and accessibility permissions your system asks for.
4. Open After Effects and start a blank composition. Every example we made began as an empty canvas.

The first time it touches After Effects you will see a permission pop up. Choose always allow, and you can watch it work while you do nothing. It even shows a second cursor moving around your screen.

What it does well

The jobs that suit it have one thing in common. They are well defined, they need several small animations that match each other, and they used to take an afternoon of keyframing.

• Titles and lower thirds. Text that builds on, holds and leaves, with easing and motion blur.
• Stat cards. A player or product card with numbers that count up, a glare that crosses it and a face that pops out of the frame.
• Animating from screenshots. Give it three or four screenshots of a website or app, describe the story, and it builds a promo with a cursor that clicks through the interface.
• Following a screen recording. Record yourself using a product and it recreates the journey with smooth zooms that keep the important area in the centre of the frame.
• Tracking graphics onto footage. It can add on-screen graphics to a clip and track them in, which used to mean fiddly camera tracking.

The part that makes the difference is how you ask. Motion designers have the vocabulary already, so they get better results. Describe the animation in order, step by step, the way you would brief a human editor. If you do not know the terms, show it a clip of a style you like as a reference.

What one piece costs in plan usage

A job usually takes about ten minutes, and ten minutes of Astra is roughly 2% of the usage on the $100 ChatGPT plan. That sounds small until you stack twenty jobs in a week. Higher effort settings are slower and use more, and for motion graphics medium or high is enough. Keep the top settings for work like colour grading.

The cheapest edit is the one you make yourself. Once the project exists, changing text, timing and colours by hand in After Effects costs nothing. Ask Astra for the changes you cannot make yourself.

Where you still finish by hand

The first pass is a starting block. On a promo we built from website screenshots, it understood the layout and animated everything, and the text positioning and the logo were still off. It did not have our logo file until we supplied one, and you should upload your own assets from the start.

It will not replace a motion designer who has done this for years. A piece our own motion designer built over about two weeks was recreated from a video reference in ten minutes. The copy was close and clearly weaker. If you already work in motion graphics, that is a way to turn jobs round much more quickly and keep the polish for yourself. If you do not, it gives you a clean base that you can learn on.

Use references to guide the style, such as your own previous work or motion you admire. Do not use them to copy somebody else's finished piece.

For example, inside GenHQ we use this to put stat cards and lower thirds over generated footage, then do the last tweaks in After Effects ourselves. The same agent can grade your footage in DaVinci Resolve too, and that job behaves very differently. It is slower and it crashes more, which is covered in can AI colour grade in DaVinci Resolve.

Common questions

Can ChatGPT make motion graphics in After Effects?

Yes. With the desktop app, Astra 6 and the computer use plugin, ChatGPT opens After Effects and builds the animation in a real project. Every layer stays editable.

Do I need to know After Effects?

Not much. Astra organises and labels the project, and if you get lost you can screenshot the panel and ask it. Knowing the basics lets you make small edits without spending plan usage.

How much does it cost?

The software is the cost of After Effects plus a ChatGPT plan. A typical job runs about ten minutes, which is around 2% of usage on the $100 plan.

Is the result good enough for client work?

As a base, often yes. The first pass usually needs fixing on positioning, brand assets and polish, so plan on finishing it yourself.

Want to learn this with people making this kind of work every week? Join GenHQ.