To make an AI cartoon video, either describe a scene and name a cartoon style so Morphic generates it animated, or turn a photo into a cartoon with style transfer, then animate that character. The decisive step is naming the exact style (anime, flat 2D, 3D, or claymation), not just asking for a "cartoon."
Steps at a glance
- Choose text or a photo to start
- Name the exact cartoon style
- Generate the animated shot
- Keep the character consistent
- Assemble and export the cut
Make your cartoon video step by step
1.
Pick your starting point
Start from text when you are inventing a scene from nothing, describing the action and the style together for the model to draw. Start from a photo when you want a real person, pet, or product to become a character, cartooning the image first, then animating it. Both routes end the same way, on the timeline; the only difference is where the character comes from. If you have a specific character in mind, starting from a photo or a fixed reference makes it far easier to keep them consistent later.
2.
Name the exact style
"Cartoon" is too vague, and the model needs the specific look, so name the style, the line quality, and the palette so every shot matches. The three prompts below produce three completely different results from the same idea, which is exactly the point: the words you choose here decide whether you get a Saturday-morning cartoon, a glossy studio render, or a hand-drawn storybook. Pick one deliberately and reuse the same style words on every shot.
3.
Generate the animated shot
Generate the clip and check that the style and motion both hold together. For a photo-based cartoon, apply the style with anime or cartoon style transfer first, then animate the stylised result rather than the original photo. Keep motion smooth and moderate: cartoon animation tolerates a little more movement than photoreal video, but heavy motion still muddies the linework and smears the colour. Generate a couple of variations and keep the one where the character stays cleanest.
4.
Keep the character consistent
If your cartoon has a recurring character, attach its image as a reference on every shot so the design carries across the whole sequence. Without a reference, each shot tends to invent a slightly different face, outfit, or proportion, and the result stops reading as one character. The reference is what turns a set of separate clips into a single character telling a story, which is the difference between a cartoon and a pile of unrelated stylised shots.
5.
Assemble and export the cut
Bring the shots onto the Compose timeline, order them, and cut on action the way animation is normally edited, letting movement carry the eye from one shot to the next. Add a music bed and any sound effects to give the cartoon energy, review the whole thing once for pacing, then export. Free exports carry a watermark; a paid plan removes it and exports at higher resolution, which matters for anything you are publishing rather than testing.
Cartoon styles and how to prompt them
The look you get is the look you name. Use this to pick a style and the words that trigger it.
| Style | Prompt cues | Best for |
|---|---|---|
| Anime | Cel shading, expressive eyes, clean linework | Character stories, action, music videos |
| Flat 2D | Bold outlines, limited palette, simple shapes | Explainers, ads, social clips |
| 3D animated | Rounded shapes, glossy light, polished | Family-friendly, product mascots |
| Claymation | Sculpted textures, soft studio light, tactile | Quirky, handcrafted, nostalgic pieces |
Starting from an idea rather than a picture? The broader how to make an AI video guide covers directing shots, and the text-to-video generator is where a styled prompt becomes an animated clip.
Put it together
A good AI cartoon video is a named style, animated with restraint, held together by a consistent character. Decide whether you are starting from text or a photo, name the exact look down to the linework and palette, and attach a reference so your character stays the same across shots. Nail the style once and every clip in the sequence inherits it, which is what makes a set of generated shots feel like one cartoon rather than five separate experiments.