Creative workflows
How to Make Motion Graphics With AI: Templates, Prompts and What Each Is For
Five motion graphics jobs a template does without AI, when to have a chat model write the scene, how to revise it, and what the free MP4 export gives you.
Video examples here were generated on Cicadas and keep their model, settings and date. Cicadas earns money when you generate with these models. How we test
Short answer: for the five jobs most people search for (steps, a big number, a bar chart, a launch teaser, a quote), fill in a template in the AI motion graphics generator: no AI is involved, the words come out exactly as typed, and the MP4 is free. For anything else, have a chat model write the scene and paste it back. Generating motion graphics as video footage is the route we would avoid, because video models misspell on-screen text.
We make Cicadas, so the tool in this guide is ours. The timings and file sizes below come from our own test runs in Chrome on a computer on October 10, 2026; the template lengths were computed from the page's code on November 5, 2026; the paid writing service was checked as switched on, at 25 credits to write and 20 to revise, on November 5, 2026.
Three ways to make motion graphics, and which one is AI
"AI motion graphics" covers three different things, and they fail in different ways.
| Route | Who decides the layout | Text and numbers | Cost on Cicadas | Best for |
|---|---|---|---|---|
| Template | A fixed layout for each of five jobs | Exactly as typed | Free | Steps, one figure, a chart, a teaser, a quote |
| Chat model writes the scene | Claude or another model, from your brief | Exactly as written in the scene; you can edit each one | Free with your own chat model; 25 credits if Cicadas writes it | Arrows, several elements per scene, your own images, up to 60 seconds |
| Video model generates footage | The model, frame by frame | Often misspelled or changing | Per clip, by model | A filmed look, not readable text |
The first two routes end in the same renderer: the page draws text, shapes, counters and bars on a canvas in your browser and writes an MP4. Nothing is uploaded for the render. The third route is a different product, covered in our prompt guide for AI video.
Pick the template by the job
| Job | Template | What you type | Length of the MP4 |
|---|---|---|---|
| Explain how something works | Steps | A title and 2 to 4 steps, one per line | 12 s with 2 steps, 19 s with 4 |
| Show one result | Big number | A label, the number, optional prefix, suffix and caption | 6 s |
| Compare a few values | Bar chart | A title and up to 8 lines such as "Jul 1.2" | 6 to 7 s |
| Announce a product | Launch teaser | A name, optional tagline, up to 3 features, a closing line | 6 s with the name only, 12 s with everything |
| Share a line someone said | Quote | Up to 180 characters and the author | 6 to 9 s, longer quotes type for longer |
Each template lays itself out again for 16:9, 9:16 or 1:1, in four colour sets. A bar label holds up to 14 characters; a step holds up to 70. In our test the Big number counter for 61,400 ended on 61,400 and stayed there, and a three-bar chart of 1.2, 1.4 and 1.6 drew the bars in those proportions, which is the part a video model gets wrong.
Make one from a template
- Open the AI motion graphics generator, keep "From a template" and pick the template that matches the job above.
- Fill in the required fields. The preview starts as soon as they pass the check; a field that is too long or a bar line without a number is marked under the box.
- Choose the shape for where the video goes: 16:9 (1920 by 1080) for YouTube and slides, 9:16 (1080 by 1920) for Reels, Shorts and TikTok, 1:1 (1080 by 1080) for feeds. Pick a colour set.
- Click Make MP4. In our October test, six template videos of 6 to 16 seconds took 1 to 5 seconds each to make; a 6-second chart at 1920 by 1080 was 0.26 MB and a 16-second steps video was 1.5 MB.
When to have a chat model write the scene
Switch to "Written by a chat model" when the template would force your idea into the wrong shape: a process with arrows between boxes, two charts side by side, a counter next to a list, your logo in the corner, or a piece longer than 19 seconds. Describe it with the exact wording and figures you want on screen, pick the shape and a length of 10, 15, 20 or 30 seconds, and copy the prompt.
Paste that prompt into Claude on any plan, or another model that can write JSON, and paste its whole answer back. The page checks every field before drawing anything and lists what is wrong if the scene cannot be drawn, so you can hand those lines back to the model. In our October test, six answers from Claude Opus, Sonnet and Haiku were all valid on the first try; the two from Haiku had text running outside its box, so we use Sonnet or Opus. We have not tested other models.
If you would rather not leave the page, "Have Cicadas write it" sends your description to a model on Cicadas for 25 credits. It usually takes 30 to 90 seconds, needs a sign-in, and returns the credits if the scene comes back unusable. A valid scene can still need layout edits.
Re-rolling and revising without starting over
People search for a way to "reroll" AI motion graphics because a generated video can only be thrown away and made again. A scene is different: it is a list of parts, so you change the part that is wrong.
- Wrong word or number: in a template, edit the field; in a written scene, open Edit the words under the preview, where every text has its own box. No model is involved and nothing is charged.
- Wrong timing, size or colour: tell the model what to change ("make the title larger and hold the last frame longer") and paste the new answer. On Cicadas a revision costs 20 credits; with your own chat model it is free.
- Wrong idea altogether: write a new brief. Up to five recent paid scenes stay in your account, so an earlier version can be applied again.
What the MP4 can and cannot be
Every route exports an H.264 MP4 at 30 frames per second with no sound, at 1920 by 1080, 1080 by 1920 or 1080 by 1080. Template videos run 6 to 19 seconds; a written scene can run up to 60. There is no transparent export and no sound, so music and voice-over go in your editor. For a thumbnail, Save this frame as a PNG stores the frame shown in the preview.
Text is drawn with the fonts on your device, so line breaks can differ slightly from one computer to another; the MP4 you download is fixed. We tested in Chrome on a computer; Safari, Firefox and phones were not part of that test, and the page says so if your browser cannot write MP4.
Smaller jobs have smaller tools
If all you need is one to three lines of text that arrive with a typewriter, word-by-word or pop effect, the free text animation maker is quicker and also exports a GIF, transparent if you like. For a logo intro or outro, the logo animation maker moves your file without redrawing it; our guide on how to animate a logo covers settings for each use. For a step-by-step explainer, the AI explainer video generator opens on the Steps template, and Claude motion graphics explains the chat-model route in detail.
Questions
Can AI make motion graphics?
Yes, in two ways. A video model can generate footage that looks like motion design, but it redraws every frame, so text and numbers often come out wrong. A language model can instead write a scene description that a renderer draws exactly. On Cicadas the free templates need no AI at all, and the AI part is the optional route where a chat model writes the scene.
Which motion graphics jobs suit a template?
Five recurring jobs: an explainer in two to four steps, one figure that counts up, a bar chart of up to eight values, a launch teaser with a name, tagline and three features, and a quote typed out with its author. Anything with arrows, several charts or your own images in one scene needs the chat-model route.
How long can an AI motion graphics video be?
Template videos last from 6 to 19 seconds, depending on how much you type: a four-step explainer is the longest. A scene written by a chat model can run up to 60 seconds; the prompt offers 10, 15, 20 or 30 seconds, and you can change the duration in the scene if you need more.
What do I do if I don't like the result?
On the template route, change a field, the shape or the colour set and the preview updates at once. On the free chat-model route, ask the model for a change and paste its new answer. If Cicadas wrote the scene for you, a revision costs 20 credits. In a written scene, any word or number can also be fixed under Edit the words, without asking a model again.
Is the export free, and does it have sound?
Rendering and downloading the MP4 are free on every route, with no watermark and no account. The MP4 is silent: add music or a voice-over in your editor, or make a sound with our sound effect generator.
Make one now
The AI motion graphics generator is free: templates, the free chat-model route, rendering and the MP4 download need no account and add no watermark. Having Cicadas write a scene costs 25 credits, and 20 per revision.