BetaAnimated scenes · video from code
Animated videos that explain, drawn on your machine in time with the voice
The CLAQON engine draws every scene with its own templates and edits it with FFmpeg: a presenter who points at what the voice says, figures that count up, timelines and objects that act. The AI model only fills in a score that the engine validates, and without a model there is a score that costs nothing.
- 14 templates and 6 signature shots
- A presenter from 50 ready-made characters
- Drawing and editing at no cost per piece
What it solves
A short that explains, without relying on stock footage
An informational video built from generic clips looks like all the others, and AI images or video cost money per piece and can show things that never happened. To explain a figure, a process or a date, what works is showing the idea.
Animated scenes turn the narration into motion graphics: every figure the voice says counts up on screen, every step of a process is ticked off as it is named and the presenter points at what matters. The six samples on this site were made this way.
How it works
From the narration to the animated layer
The voice sets the timing
Each scene starts on the word where its idea begins, and each element appears when the voice names it.
A score, not code
The account’s AI model fills in a JSON score with templates, text and icons. The engine validates it and drops anything that does not comply.
Backed figures
A figure on screen has to be in the narration or in the script’s verified claims. Otherwise it does not appear.
Measured quality
The engine scores the score (pace, variety, hook, objects that act). If the model’s score falls short of the threshold, it uses its own.
Drawing and editing
Headless Chromium, with no network, draws every frame and FFmpeg joins the layer with the voice, the effects and the music.
Capabilities
What the scenes do, in detail
- Templates
- Headline, chapter, cards, counter, bars, graph, staircase, timeline, comparison, quote, chart, fact, call to action and metaphor, plus six signature shots.
- Five styles
- Dark, paper, sketch, neon and editorial, with transitions, a drifting camera and a moving background.
- Presenter
- The account’s character presents from one side and points, nods, thinks or looks surprised on the right word. With the voice, it moves its mouth, eyes and expression.
- Objects that act
- A calculator that adds up, a stamp that approves, a traffic light that changes or a jar that fills, on the word that names them.
- Your own character
- A gallery of 50 ready-made characters and a creator with features, outfits and colours. “Create it with your AI” is optional and asks you to accept its cost first.
- Icons by meaning
- 178 Lucide icons with a concept index in Spanish and English: the icon says what is being said.
- Built-in captions
- With the spoken word marked, in four styles. The scenes keep their band clear.
- Sound on every move
- Effects synthesised by the engine itself on every appearance, with no third-party recordings.
- Scene-by-scene editing
- The editor’s “Scenes” tab changes each scene through its score and redraws it at no cost. “Regenerate with AI” quotes first.
- Video from code Off by default
- An account can produce with scenes only: no stock clips, no AI images or video. Up to 180 seconds.
Spend
What a scene costs
Drawing and editing the scenes costs nothing per piece: it runs on your machine. The only thing that can cost money is the call to the AI model that fills in the score, and that depends on your account’s model (nothing per call if it is local).
Those calls go through the single spend log, with their own budget of 5 USD a month per account. If the rate is unknown or the log is incomplete, the engine does not call the model and uses its score without AI.
Limits
Limits worth knowing
- It is in beta: the factory layer is off on every account until you turn it on.
- It is designed for vertical. Other canvases, such as 16:9 or 1:1, compose well but with more space at the sides.
- Up to 180 seconds per layer: long-form video is not produced this way.
- It draws on the CPU, with no 3D acceleration: in our tests on 20/10/2026, the layer of a 30-second short took between 30 and 45 seconds on a 4-core machine.
- It does not animate realistic characters or lip-sync real images: that is AI characters, separate and paid.
FAQ
Questions about animated scenes
Are the scenes drawn by an AI?
No. The CLAQON engine draws them with its own templates. The AI model only picks templates, text and icons in a score that the engine validates; the mode in which a model would write code is off and the Studio API never turns it on.
Can it make up a figure on screen?
It should not: the engine only lets figures on screen if they are in the narration or in the script’s verified claims, and removes the rest. That rule does not replace your review before publishing.
Do I need a paid AI model?
No. Without a model, the engine builds its own score from the narration. With a model, the score is usually richer, and it is only accepted if it passes the engine’s quality threshold.
Do I have to label the video as synthetic content?
The scenes are animated graphics, not realistic generated imagery. In CLAQON, the synthetic content disclosure on YouTube is triggered when there is generated imagery or video; deciding what to disclose is still your responsibility. Guide to labelling AI content.
Can I use my own character?
Yes. Pick one of the 50 ready-made characters or create your own with features, outfits and colours. The “Your face” photo never leaves your browser.
Available now
Animated videos for your channel, made by us
We make them for you with the factory, starting with a 10-video pilot on your topic. If you would rather have CLAQON on your own machine, pre-register.