Making cartoon videos that work on Shorts, TikTok and Reels

A cartoon that plays beautifully on a laptop can die instantly in a vertical feed. Not because the animation is worse, but because a feed is a different medium with different physics: no chosen start, no sound by default, a thumb hovering over the next video, and a third of the screen covered by interface.

Everything below is about that gap. The animation skills carry over. The framing, pacing and finishing do not.

Frame for 9:16 from the start, never by cropping

The single most common mistake is generating a 16:9 cartoon and cropping it to vertical afterwards. It fails because the model composed the shot for the frame it was given — the character sits at a third, the action reads across the width, the background fills the sides. Crop that and you get a character half out of frame with the interesting part removed.

Set 9:16 before generating. And know the constraint that catches people out: when you supply a source image or video, the output inherits that asset's aspect ratio, not the one you selected. If you are animating a photo for a vertical feed, crop the photo to 9:16 first.

Respect the safe area

Platform interface covers a serious amount of a vertical frame: roughly the top tenth (profile, sound), the bottom fifth (caption, handle, music ticker) and the right edge (the action rail). Anything important in those zones is obscured.

Practically, keep the subject and any key action in the middle 60% of the height and away from the right edge. Prompting for "centred composition, subject fills the middle of the frame, empty space at top and bottom" gets you most of the way there.

The first two seconds are the entire job

In a feed, the viewer has not chosen your video — they have been handed it, and they are deciding in about two seconds whether to keep it. Nothing else in the clip matters if that fails.

Three things that work in an animated opening:

Open mid-action. Not a character walking into frame — a character already doing something inexplicable. Establishing shots are a cinema convention and they are lethal here.

Open on a question. Something visible that the viewer cannot immediately explain. Why is that robot holding a birthday cake in a snowstorm?

Open on the unusual style. In a feed dominated by camera footage, a strongly drawn frame is itself a pattern interrupt. This is the structural advantage animation has in short-form, and it argues for a distinctive style over generic flat vector — the differences are covered in our styles guide.

What does not work: a logo, a title card, a slow fade in, a character introducing themselves. All of these spend the only currency you have.

Assume the sound is off

Most short-form video is watched muted, at least at first. This has two consequences that people acknowledge and then ignore.

The visual has to carry the meaning alone. If your joke is in the dialogue, it does not exist for most of your audience. Test it by watching your own clip on mute — if you cannot follow it, neither can they.

Captions are mandatory, not a nice-to-have. Burn them in rather than relying on platform auto-captions, which are inconsistently displayed and often positioned badly. Keep them large, high contrast, in the middle band of the frame, and no more than about six words at a time.

None of which means turn the audio off when generating. Synchronised sound noticeably lifts the experience for the minority who do have sound on, and platforms measure their engagement too. Generate with audio, then add captions on top.

Length: shorter than you think

Thirty seconds is available and it is often the wrong choice for a feed. Completion rate is the metric these platforms respond to most, and a clip that says its piece in twelve seconds and ends will outperform the same idea stretched to thirty.

Design the loop deliberately

Short-form players restart automatically, and a rewatch counts as engagement. A clip whose last frame flows into its first gets watched two or three times by the same viewer without them consciously deciding to.

The clean way to build this is first-and-last-frame generation: supply the same image as both the opening and closing frame, and the model animates a journey that returns to where it started. The join becomes invisible.

Even without that, you can help the loop by ending on the same composition you opened on, and by avoiding an ending that resolves too firmly — a character settling into stillness signals "over" and breaks the cycle.

Make variants, because that is now cheap

The old economics of animation forced you to bet everything on one execution. That constraint is gone, and short-form is precisely where that matters, because feed performance is genuinely hard to predict.

A sensible pattern: make one clip, then make three alternative openings for it. Same body, different first two seconds. Post them across a week. The winner is frequently not the one you expected, and the gap between best and worst opening is usually larger than the gap between a good and a mediocre animation.

A finishing checklist

• Generated at 9:16, not cropped to it.
• Subject inside the middle 60% vertically, clear of the right edge.
• Something inexplicable or in-motion within the first half second.
• Watched on mute end to end and still comprehensible.
• Burned-in captions, large, high contrast, mid-frame.
• Any held frames trimmed off the head of the clip.
• Ending that flows back to the opening, if a loop suits the idea.
• No rendered text inside the animation itself — added in the editor instead.

Where to start

Pick a concept with a built-in question — the escalating loop and the slow reveal from our ideas list are the most reliable in a feed — set 9:16 and about ten seconds, and generate it in the cartoon video creator. Then make the same clip twice more with different openings. That third version is usually the one that travels.