How to make AI 3D animated ads

What is an AI 3D animated ad?

An AI 3D animated ad is a short ad in which the characters and the world are stylised 3D animation made with AI image and video models, not filmed. It usually tells a small story (a problem, a turn, the product, the result) with cartoon characters and soft, warm light.

The look most people search for is "Pixar style": rounded shapes, large eyes and a polished family-film finish. Public examples of the format are on Pixar style AI ads, and handmade-looking variants are on AI claymation and craft ads. Those examples come from posts on X; they were not made on or for Anglr.

The format stands out because it does not look like phone footage. Peter Quadrel describes the wider category as a real product placed inside a generated world, in styles from claymation and diorama to plush toys and 3D.

Tools creators use for it

The table counts how many of the reviewed tutorials named each tool for animated ads. A tutorial that used two tools counts once for each.

Based on 12 public tutorials posted between 9 May 2026 and 11 Oct 2026, reviewed by Anglr.

Single tutorials also used Veo as another video model, Arcads (through its MCP connection to Claude), MaxFusion to carry the work from research to variants, Gemini for the script, Google Flow for animation and lip sync, Arena.ai for stills, and CapCut for the edit. The sample is small: the 3D style has fewer published methods than UGC, and several posts stop before the full method.

How do you make a 3D animated ad with AI?

You make it still first: look, script, character sheet, start frames, animation, sound and cut. The order below is our synthesis of the tutorials, with notes on where they agree and where they split.

1. Lock one look and describe it the same way every time

Choose the style before anything else and write it down as a short, fixed description. At least four authors stress this. Peter Quadrel locks a style first so new variants are quick. @maxxmalist asks for one named family-film look for every character sheet. @Zarnab_with_Ai and @Zyrellix put the style at the very top of the prompt.

Chibueze Kingsley explains why it matters most: "Lock your core descriptor tokens before you generate a single frame." If the words that describe the character or style drift between scenes, the face glitches once it moves.

Pick the angle with the look. Animation can show what a camera cannot, and Muhammad Basit lists the options he uses: the inside of the product, the product as a character, the customer's problem drawn as a creature, or a transformation you could never film. Choose one, so the ad has a single reason to exist.

Many creators write a studio name into the prompt. We suggest describing the look instead (stylised 3D, rounded forms, oversized eyes, soft global light, warm palette). It gives you more control, and the finished ad should never use a studio's name, characters or logos.

2. Write and lock the script before you animate

Finish the script, including every spoken line, before any video credit is spent. Hamza Malik has Claude write it from a positioning research prompt. Chibueze Kingsley writes it in Gemini around a hook built for a phone screen. @maxxmalist uploads the finished script and has GPT adapt a prompt template to it. Romain Torres gets the voiceover written first, then the clip prompts.

The reason is cost. Chibueze Kingsley warns that changing dialogue after generation can cost a whole day, because every affected scene has to be generated again. Animated ads are story-led, so a plain structure helps: who the character is, what goes wrong, how the product fixes it, what life looks like after. Muhammad Basit writes his 30-second story as timed shots, with the hook in the first three seconds, the problem, a demo or transformation, the payoff and the brand at the end, and he starts from customer reviews, objections and competitor ads before writing anything. Our UGC script template gives the hook, problem, product and call-to-action beats in more detail.

3. Build a character sheet (and the rest of the world)

Make a sheet for each recurring character: the same character from several angles, in the locked style, on a plain background. @maxxmalist starts from realistic photos and asks for stylised character sheets from them. Chibueze Kingsley fixes one consistent character and scene before generating. @OriSilver uses contact sheets as the backbone of the method, so the video model follows the planned shots.

Muhammad Basit goes further and builds three sheets before any frame: character (front and side views, expressions), product (materials, packaging) and environment (lighting, props, colour). Attach the sheets to every later generation. If your product appears, make a clean product reference as well, so the label and shape stay the same in a cartoon world. Peter Quadrel's approach keeps the real product and generates only the world around it.

4. Generate one start frame per scene

Make each scene as a still first, in the image model, with the character sheet attached. Hamza Malik makes a stylised start frame per scene in Nano Banana Pro. Chibueze Kingsley animates stills rather than prompting video from text.

Muhammad Basit storyboards six to ten decisive shots and approves them before generating any still, then makes start frames, end frames and transition references in GPT Image 2 and Higgsfield. Stills are where you fix the look, the framing and the product. Check every frame side by side for the same face, proportions and colours before you animate any of them.

5. Animate each scene from its frame

Animate one scene at a time from its start frame. Hamza Malik animates each scene in Kling 3.0. Chibueze Kingsley animates in Google Flow, which also handles transitions and lip sync in that workflow.

Muhammad Basit directs the motion in Seedance 2.5 shot by shot, giving each one its camera, action, timing, transition, sound and performance.

Two authors work differently. @Zarnab_with_Ai and @Zyrellix write one long text prompt in Seedance and let it produce the whole short story in one run, with the style, the character and each beat spelled out. That is faster but gives less control over each shot. Romain Torres sits in between: one prompt per timed clip, rendered in Seedance 2.0 through the Arcads connection in Claude.

Kling's own image-to-video guide says to describe the subject and its movement, because the image already sets the scene. A description far from the image can cause a cut. Keep motion prompts short and about the action.

6. Add voice and sound, then cut

Finish in an editor. Hamza Malik joins the clips in CapCut with speed ramps and then upscales. Romain Torres lays the voiceover over the finished animation. Neither uses the raw generation as the final ad.

Sound carries much of an animated ad: a narrator or character voice, music under the story and small effects on each action. Keep one voice for each character across the whole ad.

Remake a winning ad in 3D

Once an ad works in one format, the same script can run again as animation. @maxxmalist took the character from a winning avatar ad and gave Seedance 2.5 its image, video and audio references to remake it as a 3D cartoon. @Zyrellix crosses formats the other way, writing a 3D cartoon makeup tutorial as a slow, close-up ASMR piece. Both keep the tested message and change only the look, which makes the test a fair one.

Optional: mix live action and animation

@techhalla opens on a handheld live-action room, hard-cuts into a hand-inked cartoon chase, then cuts back to the room. The switch in look is the hook. It is one author's method, but it is a simple way to make a product's "before" and "after" feel like different worlds.

Prompt template for a 3D animated scene

These templates are ours. The start frame follows Google's Nano Banana guide (reference images, how they relate, then the new scene, with the purpose of the image stated). The motion prompt follows Kling's image-to-video advice: describe who moves and how, and let the image carry the scene. Duration and aspect ratio go in the settings. Build your own in the AI ad prompt builder, or see Kling prompts and Nano Banana prompts.

Character sheet (image model):

Using the attached photo only as a guide for [two or three features: hair, build, clothing], create a character sheet for an animated ad: stylised 3D animation, rounded forms, large expressive eyes, soft warm global light, [palette]. Show the same character front, three-quarter and side, full body, neutral pose, on a plain light grey background. Keep every view identical in face, proportions and colours.

Start frame (image model, per scene):

Image 1 is the character sheet. Image 2 is the product. Place the character from Image 1 in [setting], [time of day], doing [one action] with [product] from Image 2 clearly visible and its label readable. Same style as Image 1: stylised 3D animation, rounded forms, large expressive eyes, soft warm global light. Framing: [shot size], [camera angle]. This is a frame from a short animated product ad. Vertical 9:16.

Motion (Kling, image-to-video):

The character [one action, small and slow: lifts the jar, turns to the camera, smiles]. Their hair bounces slightly as they move. The camera pushes in slowly.
Character (bright, friendly): [one short line from the script]
Gentle music, [one small sound effect on the action].

Common problems and fixes

What creators report

These are the authors' own claims, not results we checked.

• Hamza Malik reports that animated creatives pull higher hook rates than standard UGC across several accounts, and that one ad takes under two hours.
• Peter Quadrel reports that with the style already fixed, each new variant takes about 30 minutes.
• @maxxmalist reports making an animated ad in under 10 minutes with his character-sheet and template method.
• Romain Torres reports that his Claude skill turns a week of motion design into about five minutes.

If you already make animated ads that perform, the creators page explains how Anglr pays for them. The AI commercials guide covers live-action brand films, and the AI UGC ads guide covers the selfie format.