Animated Explainer Video Skill for AI Agents

The whole thing

One line in. A finished explainer out.

Nothing below was edited by hand. This is what came back from a single line for a real brand, with the story, the narration and the score the skill wrote itself.

Make an animated explainer for Arata's hair growth serum. Papercraft, 16:9.

Start creating Install the skill

Your brand, your topic, any of the ten looks.

Just ask

Say the word, it runs.

The skill fires on natural requests like these. No commands to memorize.

“Make an explainer video about X”“Claymation / pixel art / whiteboard explainer”“Explain my product in a short animated video”“Turn this topic into a narrated explainer ad”

The ten styles

One pipeline. Ten looks.

Name a style and the skill loads that cartridge: the medium, the cut rhythm, the typography, the camera language and the sound all change together. Every clip below was made with it.

Isometric Flat Vector

Monochromatic high-contrast block-print illustration, heavy distress texture, stark white focal glows, tense motion-comic cuts.

Best for: Tech, security and dev-tool brands. Dramatic reveals, strategic planning analogies, tense storytelling, cybersecurity concepts.

2D Illustrator

Lineless digital gouache on twos, toothy paper grain, hard geometric shadows on a flat saturated color void, lo-fi cool.

Best for: Lifestyle, sport and edgy tech brands. Fast-paced instructional content, product tutorials, and manifestos that need high-energy visuals under a calm, authoritative voice.

Whiteboard Doodle

Dry-erase marker doodles on a smudged whiteboard, stroke-by-stroke draw-ons, line boiling, squeaky-marker foley.

Best for: Education, how-it-works and onboarding topics. Classic explainer feel, approachable, ideal for simplifying complex abstract concepts and quick trivia hits.

Pixel Art

Vibrant 16-bit arcade pixel art, chibi sprites, deep parallax, snappy on-twos sprite motion, chiptune.

Best for: Gaming, tech and cybersecurity brands. Gamified explainers, nostalgic brand activations, and high-energy educational shorts.

Claymotion

Vibrant plasticine stop-motion, visible fingerprints, chunky characters, sculpted clay text, 12fps boil.

Best for: Playful, whimsical, story-driven brands. Tactile brand storytelling, educational explainers, whimsical product introductions.

Fluffy Toy

Cozy needle-felted wool diorama, fuzzy stray fibers, bead eyes, physicalized yarn VFX, gentle stop-motion.

Best for: Warm, gentle wellness, kids, health and food brands. Explaining gentle, complex or emotional topics, and brand storytelling built on warmth and approachability.

3D Mix

Soft pastel bloom illustration, limbless kawaii blobs, drifting sparkles, gentle squash-and-stretch, lullaby audio.

Best for: Wellness, mental health and cozy consumer apps. Gentle educational explainers and soft product introductions.

Low Poly

Warm low-poly 3D world, un-smoothed flat-shaded facets, smooth camera glides, childlike acoustic storybook.

Best for: Kids education, eco-friendly products, and brand stories about simplicity, warmth and grassroots origins. A clean, modern, slightly premium storybook.

3D Papercraft

Theatrical die-cut papercraft diorama, deep parallax tunnels, jewel-tone rim light, warm storybook narration.

Best for: Whimsical, magical brand storytelling. Historical or mythological explainers, holiday campaigns, and product origin stories.

Mixed Media

2.5D paper-cut collage, greyscale photo cut-outs with white scissor borders, hard drop shadows, sticky-note infographics.

Best for: Education, data and mechanics topics. Explaining hidden mechanics, physics, data-heavy comparisons and historical timelines where physicalizing an abstract concept aids comprehension.

The pipeline

What the skill does for you.

1
Pick a style cartridge
The skill recommends one of ten looks from your brand and topic, then shows the full menu so you can override by eye. A style is a whole art direction: the medium, the cut cadence, the camera language, the typography and the audio all travel with it. Unsure between two? It bakes the first keyframe in each and lets you choose.
2
Approve the beat map
It invents an original concept for your topic (never a style file's example), picks a narrative arc, and lays out every shot: framing, camera move, element motion, any baked headline, and the narration line, timed to that style's words per second. This is the mandatory gate.
3
Render the keyframes
One still per shot, with the style's image block travelling verbatim into every prompt (gpt-image-2 for the spec-faithful styles, nano-banana-2 for the organic textures). That verbatim block is what makes eight shots read as one film. Off-style frames get re-rolled here, where they are cheap.
4
Animate every shot
Each approved keyframe becomes a clip carrying the style's motion DNA on Gemini Omni Flash, or Seedance 2.0 when a shot needs an exact end-frame landing like papercraft's z-axis tunnel. Clips are SFX-only, so nothing collides with the voice.
5
Snap the stop-motion styles
The five on-twos styles read as roughly 12fps, and no video model holds that on its own. Each clip gets an ffmpeg step-frame pass before the mix, which is what actually sells the claymation and felt as stop-motion.
6
Narrate, score, assemble
One voiceover line per shot, each pinned to its own start time (a single continuous read drifts off the cuts), plus one instrumental bed. A single render_composition call assembles the clips, the narration and the score, ducking the music under the voice automatically.

Ten looks. One pipeline.

Pick the style, approve the beat map, and the agent does the rest.

Start creating

Your product, in the medium

It doesn't paste your product in. It rebuilds it.

A photo dropped onto a clay set is the fastest way to break the spell. So on the shots that feature your product, the skill passes its photo as a reference and remakes it as felt, paper, marker or pixels, keeping the silhouette and the label readable.

Your product photo becomes felt, paper, clay or a doodle, and the label stays legible
Mixed-media collage keeps a real greyscale photo cut-out, because that is the style
No hero product? A pure topic explainer works too, with no product beat

Start creating Install the skill

Som Sleep

Arata

Mixed Media

Whiteboard Doodle

The model stack

The right model for each style.

The style file picks the defaults, one image model and one video model per film. GPT Image 2 draws the precise looks, Nano Banana 2 the textured ones, Gemini Omni Flash moves nine styles out of ten and Seedance 2.0 takes the papercraft tunnel. Ask for the newer ones, GPT Image 2.5 Sunburst, Gemini Omni Flash 1.1 or Seedance 2.5, and the agent runs them on every shot instead.

What you get

A finished narrated explainer video (16:9 or 9:16), assembled and mixed
One keyframe and one clip per shot, filed separately in your library
A per-shot narration track and an instrumental score
An editable composition that reopens in the Advibly video editor

Start creating Read the source

GPT Image 2.5 SunburstNew

On request, sharper keyframes for the precise styles

GPT Image 2

Keyframes for the precise styles, from whiteboard to pixel art

Nano Banana 2

Keyframes for clay, felt and gouache, where texture is everything

Gemini Omni Flash 1.1New

On request, the newer Omni for the same nine styles, up to 1080p

Seedance 2.5New

On request, one shot up to 30 seconds that still lands on an exact end frame

Gemini Omni Flash

Motion for nine of the ten styles, SFX only so nothing talks over the voice

Seedance 2.0

Papercraft's deep tunnel, which has to land on an exact end frame

xAI TTS

The narrator, one line per shot so the voice stays on the cuts

MiniMax 2.6

The instrumental score, ducked under the voice automatically

Install

Add it in seconds.

Install the skill, then make sure the Advibly MCP is connected so it has tools to call. Works with every skills-capable agent.

Claude Code

Claude

Cursor

Codex

VS Code

OpenClaw

Hermes

ChatGPT

Connect the Advibly MCP first

One click, no terminal

Download .skill Download .zip

Drop the .skill file into Claude → Settings → Skills, or unzip the folder into ~/.claude/skills.

One command (any agent)

npx skills add advibly/skills -s explainer-videos

Works with Claude Code, Claude Desktop, Cursor, and any agent that supports skills. Pulls straight from GitHub.

Manual (Claude Code)

git clone https://github.com/advibly/skills.git
cp -R skills/explainer-videos ~/.claude/skills/

Or in Claude Desktop: Settings → Skills → Install and select the skill folder.

Bundled references

The playbook.

Each skill ships with reference docs the agent loads on demand, so the craft is baked in.

style-catalog.md

The ten-style menu with one-line signatures, the recommend-from-brand logic, and the model / cadence / voice / text routing table.

styles/<style>.md

The ten cartridges. Each holds the full look, motion, typography and audio DNA plus paste-ready image and motion prompt blocks and its failure modes.

story-and-beats.md

The narrative arc library, hook patterns, shot cadence per style, the beat-map schema, and the anti-monotony checklist.

pipeline-and-audio.md

The asset workflow, the image and video model matrix, the on-twos step-frame snap, per-shot voiceover timing, and the composition call.

Honest limits

What to watch for.

One style, one image model, one video model per delivered video. The set swaps whole or not at all; mixing styles inside one explainer is what breaks it.
The look is born in the keyframe. If a still is not convincingly in-style, no amount of motion saves it, so off-style frames get re-rolled before animating.
The stop-motion cadence comes from the step-frame pass, never from the video model. Without a shell to run ffmpeg, those five styles ship smooth.
Baked headlines are rendered by the image model, since there is no text-overlay tool. Only five of the ten styles carry in-world text at all.
Real people or third-party marks move the video to Kling; the default motion models block recognizable faces and logos.

Questions

Before you install it.

Which styles can it make?

Ten: claymation, felted wool, pixel art, whiteboard doodle, papercraft shadowbox, low-poly 3D, cozy kawaii, textured gouache, a cinematic block print, and mixed-media collage. The skill recommends one from your brand and topic, then shows you all ten so you can pick by eye. One style per video, never mixed.

What do I give the agent?

A brand already set up in Advibly and one topic or claim to explain. A product photo is optional; a pure topic explainer works too. It asks for 16:9 or 9:16 (16:9 by default), writes an original concept, beat map and narration, and shows them to you before a credit is spent.

How long does a full explainer take?

The gates are the slow part, not the models. You pick the style and approve the beat map. After that the keyframes, clips, narration and score run in the background, and the finished cut lands in your Advibly library with a link to fine-tune it in the video editor.

What does it cost?

The skill is free and open source. Running it spends your normal Advibly credits on the images and clips it generates, with no surcharge for the skill, and every generation shows its credit cost before it runs.

Which agents can run it?

Any agent that supports installable skills and the Model Context Protocol: Claude Code, Claude, Cursor, Codex, VS Code, OpenClaw, Hermes and ChatGPT. Install with one npx command, then connect the Advibly MCP so the skill has tools to call.

Does the claymation actually look like stop-motion?

Yes. Video models always render too smooth to hold 12fps, so the skill never asks them to. The stop-motion styles get an on-twos effect on the final composition instead, the same effect you can toggle under Effects in the Advibly editor.