Make an animated explainer video in this order: decide what the viewer should believe when the video ends, cut that belief into scenes with one idea each, build the visuals, review still frames before you pay for any motion, and write the narration last, to the finished picture.
Most first explainers get made in the reverse order: someone writes a voiceover, records it, and hands both to an animator. Pictures bolted onto fixed words can only illustrate them, so the result is a narrated slideshow. Nearly every rejected video in our production history was made words-first.
Petal, the invoicing product in the examples below, works like this: you draft an invoice and send it, Petal tracks it from sent to paid, and anything past due surfaces on an overdue list. The method is the same whether you build the video yourself or brief someone else.
Start from the change in the viewer's head
Asking what the video should show produces a tour. Asking what the viewer must be able to think after watching produces an explainer. Plan in this order:
- Learning outcome. What can the viewer think afterwards that they couldn't before?
- The one idea. A single thesis sentence.
- The causal chain. Orient from what they already know, then macro, then micro, then one worked example.
- Beats. The chain cut into scenes, one idea each.
- Visuals per beat. Described in words before anything is built.
- Narration. Written last, to the finished picture.
For Petal: the outcome is "invoices track themselves, and anything overdue surfaces in one list." The one idea: Petal chases invoices so you don't have to. The chain runs from drafting invoices by hand, to sending one and watching it flip into a tracked state, to a payment marking it paid untouched, to the stragglers collecting on one overdue list. Four beats.
If a beat has no visual, the beat isn't understood yet. A scene you can't picture is a scene you haven't finished thinking about. Noticing that on paper costs nothing. In a finished render the same catch costs a rebuild.
Write the one idea as a sentence
Every script we produce opens with the one idea: a single sentence everything in the video exists to make feel obvious. Two tests separate a claim from a topic:
- The sentence must be falsifiable. "Petal makes invoicing effortless" fails, because no viewer could check it. "You send an invoice once and Petal tracks it to paid without you touching it again" passes.
- Everything on screen must serve it. Any scene that doesn't push the sentence toward obvious gets cut, however good it looks.
Each scene then gets its own thesis line. The Petal scenes are named by their
ideas (send-flips-state, paid-marks-itself, overdue-surfaces), and a
scene you can't name that way carries more than one idea
(one idea per scene).
Show the real product, or don't show a product
The accepted videos in our review history were built from the product's real shipped surfaces; the rejected ones invented their content (the built-twice comparison).
The cheap protocol: before scripting, have someone with a live account build the exact demo the video will show, run it once, and send the configuration, the output, and the run log. For Petal that means creating a real invoice, sending it, paying it through the test gateway, and leaving one overdue, so every value on screen traces to a record. A value you can't ground stays off screen, because a user who knows the product spots a fake one instantly.
Lock the scene list before anything gets built
The scene list is binding: once approved, production builds exactly those
scenes, and changes arrive as new versions, never silent edits. Each scene
carries a name that states its idea (overdue-surfaces, never scene-4), a
duration (6–8 scenes of 8–12 seconds is the working range), a visual
described concretely enough to build from, a one-line beat intent for
whoever writes narration later, and a boundary note, because a scene's exit
state must equal the next scene's enter state. The row-by-row anatomy, with
a worked contract-grade entry, is in the
storyboard guide.
Review stills before you pay for motion
Production splits into two phases. Phase one builds only the static set piece and renders stills (the money shot, the final settled frame, one per camera framing), then stops. Motion, narration, and final render wait for the stills to pass.
One batch skipped the gate: nine of ten videos were rejected, mostly for failures fully visible in a static frame, like wrong layouts, invented UI, and cropped set pieces. The next cold batch, with the gate in, passed first review five out of five. Everything expensive comes after decisions a still already exposes, so one look at one frame catches that failure class at roughly 5% of a finished video's cost.
Two more habits came out of those reviews:
- Review every camera framing, not just the hero frames. Two builds shipped croppings that cut the set piece at both edges, because only the money framing had been checked.
- Accept only images, never assurances. "Looks good on my end" is not a still, so reviews come back as rendered frames.
The same gate logic works outside video: approve the cheapest artifact that exposes the most expensive failure class.
Write the narration last
Narration gets written to the locked visuals, for three reasons:
- Sync only works in one direction. Words written to a finished picture can land the key word on the key visual moment. Visuals animated to fixed narration never quite land, and read as floating.
- The voice and the picture divide the labor. The picture shows state, and the narration says what the picture can't: the why, the name, the caveat. Never put a sentence on screen that the voice already says.
- Cheap edits stay cheap. A scene's final length is the larger of its visual minimum and its audio plus a breath pad, so a wording fix re-times itself and never touches the expensive picture.
For paid-marks-itself, the locked visual shows the invoice row flipping to
paid as the payment arrives, so the narration says the one thing the picture
can't: "You didn't touch anything. The payment came in, and Petal marked the
invoice itself." A narration voice reads about 1.9 words per second
(measured in the voice over guide), so a
9-second scene holds one short sentence; real rejected-and-accepted lines
are in the script guide.
Budget for revision
The first-pass build is about half the real effort. Across our production records, simple videos took a handful of revision passes, the flagships were re-made as full staged versions, and roughly half of all production work happened after first review (the revision numbers). Two habits keep the revision half cheap:
- Never overwrite a version. Keep every take renderable, so rejecting one costs nothing and comparing two costs a look.
- Restage a failed video instead of patching it. The videos in our record that finally passed were rebuilt, not touched up.
FAQ
Can I write the script first if writing is how I think?
Write the beats and the beat intents first; that part is thinking, and it belongs early. The narration prose waits until the visuals are locked, since the exact words carry the timing.
How long should the video be?
Sixty to ninety seconds for a single concept, which at 6–8 scenes is where tight videos land (the runtime math).
What if I can't describe a scene's visual?
The beat isn't understood yet, and you found out before spending anything. Fix the thinking, and the drawing follows.
Do I need to be an animator to use this method?
No. Every gate in it, from the one-idea sentence to the scene contract to the still review, is a planning and judgment step. Starting from zero on the production side, begin with how to make animated videos.
You can skip the brief entirely: send your product's URL and pick from twenty short candidate videos.