GUIDE
How to make viral animated videos
A 15-second vertical animated short in three steps: the premise, a 12-panel storyboard, one video prompt.
The hard part is not finding an idea. The hard part is keeping the same face, the same outfit and the same room across all twelve frames. The pipeline below exists for exactly that: each step produces the input for the next, and no step gets skipped.
What you end up with
- Duration
- Exactly 15.00 seconds
- Format
- Vertical 9:16
- Frames
- 12 panels
- Audio
- No dialogue
Three steps
The premise
Pick one single event: a hook visible in the first two seconds, a fast cause-and-effect chain, and an ending that closes inside fifteen seconds. If you have a video you like, upload it — its structure and pacing become the reference while the characters and the location are made original.
- What you get
- A one-paragraph production summary: hook, event flow, ending and dominant emotion.
- The rule that holds
- An idea that needs dialogue to be understood does not work in this format. Neither does on-screen text.
Characters and storyboard
Design the fewest characters the story actually needs and lock each one into a table: face, hair, build, outfit, footwear, behaviour. Then split the story into exactly twelve panels, each carrying one readable action.
- What you get
- A character map, image prompts for every character you don't already have, and a separate panel prompt for each of the twelve panels.
- The rule that holds
- The panel durations are added up and must come to exactly 15.00 seconds. Action panels run 0.7–1.1 s, reactions and the ending 1.2–1.8 s.
The video prompt
Turn the twelve panels into one timestamped English prompt for the video model. Every line is a shot: a start pose, one clear action, an end pose and a visible emotion.
- What you get
- The reference upload order (@image1, @image2, @image3), the master prompt and a short quality checklist.
- The rule that holds
- The continuity lock and the negative constraints stay inside the prompt. The storyboard is a shot guide only; it never appears in the video.
The real work: the identity lock
Facial proportions, hair, body proportions, outfit construction and colours, footwear and fixed accessories are identical across all twelve panels. Location layout, lighting direction, prop positions and the character's screen direction never change without a reason. That is what the lock paragraph at the end of the character prompts and the video prompt is for — carry it verbatim, never shorten it. Identity drift is the single mistake that ruins a video in this format.
Where to start
The whole pipeline is packaged and ready: the rules of all three engines, the tables and the lock paragraphs are inside it. Upload the pack to Claude, turn it into a Gemini Gem, paste it as a single no-setup prompt, or reach it through the library connector — all four run the same text.
Common mistakes
- Cramming two events into one panel. If it splits in two, it already is two panels.
- Eyeballing the timing. A storyboard isn't finished until the total has been computed.
- Trying to carry the scene with camera movement. Readability comes before motion.
- Summarising the lock paragraph. Every line you trim is a consistency gap.
- Feeding the storyboard collage as a reference and letting the grid show up in the video.
- Copying a copyrighted character or a real person. Take only structure and pacing from a reference video.
Where you run the prompt
The character and panel prompts go to an image model; the master prompt goes to the model that renders the video. That step doesn't happen on this site: you copy the prompt and run it in your own generator.
Some links on this page are affiliate links.