Free guide · 7 Nov 2026

How to make animated dialogue videos for Shorts and Reels

How to make two-character animated skits for Shorts and Reels: the five-beat script, two voices, sitcom staging, captions and posting, with an example script.

By John Shemron, graphic designer and video editor at PromptVault

An animated dialogue video is a short where two cartoon characters talk through one small, relatable mistake. To make one for Shorts, Reels or TikTok, you write a script in five beats, give each character a distinct voice, stage it in close-ups like a sitcom, add big captions, and keep it to about 45 seconds. This guide walks through every step, with a full example script.

Why do animated dialogue videos work on Shorts and Reels?

One character talking to camera is a lecture. Two characters talking is a scene, and a scene has conflict: one wants to do something, the other knows it will go wrong.

The format has a few things going for it on Shorts and Reels:

It reads on mute. Most of the joke lives in the captions, so it still lands for people scrolling with the sound off.
The viewer is the third character. They usually know the outcome before the hero does, and that gap is what keeps them watching.
A fixed cast builds a following. People come back for characters they know. Episode ten is easier to sell than episode one.
It is cheap to repeat. Once the two characters exist, each new episode is mostly a new script.

What format should an animated dialogue video use?

	What works
Length	40 to 50 seconds
Size	9:16, 1080×1920, 30 fps
Cast	Two characters: one who wants to do the risky thing, one calm voice of reason
Story	One bad idea your audience recognises
Script	12 to 16 short lines, none longer than about 8 words
Captions	Big, centred, 2 to 5 words at a time
Ending	A last line of three words or fewer

Step 1: Pick one bad idea your audience recognises

The best ideas are things your viewers have done themselves and know they shouldn't: the shortcut, the impulse buy, the "it'll be fine". Write each idea in one sentence with this shape:

A wants to do X. It goes wrong because Y. The last line is Z.

Ideas to adapt to your niche:

Shopping: buying something because "the sale ends tonight" (it ended last week too).
Fitness: skipping the warm-up "just this once".
Money: checking the budget after the purchase.
Work: "one more meeting" to save time.
Marketing: posting without proofreading the caption.
Developers: deploying on a Friday afternoon.
Students: starting the essay the night before.
Parents: "five more minutes" of screen time.
Creators: deleting a video because it flopped in the first hour.
Security: writing the password on a sticky note.

Tip: Your comments and DMs are the best source. Any complaint you hear more than twice is an episode.

Step 2: Write the script in five beats

Short comedy needs structure more than jokes. This five-beat shape works for almost any bad idea:

Beat	What happens
1	A states the plan in the first line. B answers with one short fact that should stop them.
2	A past disaster comes out in pieces, with a number that grows.
3	A tries a sensible fix, and it turns out A caused the problem.
4	A does it anyway. The disaster repeats.
5	B gets the last word, in three words or fewer, calling back to beat 2.

Here is a full example in the shopping niche. A is the shopper, B is the calm one.

Beat	Lines
1	A: "Add to cart. Sale ends at midnight." / B: "It's the same sale as last week."
2	A: "This one's different. It's an air fryer." / B: "You said that about the air fryer." / B: "And the second air fryer." / B: "Three. You own three."
3	A: "Fine. Check my budget first." / B: "You're forty over." / B: "Since the air fryers."
4	A: "...Buy it anyway." / B: "Order placed. One air fryer."
5	B: "Four. Congratulations."

Why it works:

Line one is the hook. The viewer knows the plan and the problem within three seconds.
The number grows (one, two, three, four), which gives the viewer something to track.
Beat 3 flips the blame. The fix exposes the real cause, and that is usually the biggest laugh.
The last line is a callback. It reuses the running count, so it pays off without explaining anything.

Read the script out loud with a timer. If it runs over 50 seconds, cut lines from beat 2 first.

Step 3: Cast two voices that contrast

The voices do half the characterisation. Make them different in pitch, pace and energy: A is quick and defensive, B is slow and flat. A deadpan voice of reason is funnier the less it reacts.

Three ways to get the voices:

Option	Good for	Watch out for
Record them yourself	Full control, free	Two convincing voices from one person takes practice
A paid AI voice service, such as ElevenLabs	Thousands of ready-made voices, quick	Pay-per-character credits; a 45-second script is roughly 900 characters
A free voice model on your own computer	No per-episode cost, private	Setup, and a computer that can run it

For the free options, their licences and three ways to make the mouths move, see free AI voices and lip sync for cartoon characters.

Whatever you pick, keep each character's voice identical across episodes. A new voice breaks the character as quickly as a new face does (for the face side, see how to keep the same character in AI animation).

Timing matters as much as casting. Leave a short pause after B's fact in beat 1, and a longer one before the last line.

Step 4: Stage it like a sitcom

You don't need complex animation. Short dialogue videos are mostly close-ups:

Cut to whoever is speaking, and hold a reaction shot on B after A's worst lines.
Add a slow push-in on every shot. A camera that creeps forward keeps static characters feeling alive.
Show the joke on a screen. An insert of a clock, a receipt or a budget app makes the number in beat 2 visible.
Keep the world simple. One location, one colour palette, the same layout every episode.
Move the mouths with the voice, and give B one fixed "flat" expression for deadpan lines.

The harder question is how to get two characters moving at all if you don't animate. There are three common routes (the prompt animation guide covers how to write the prompt for each):

Route	How it works	Strength	Weakness
AI video model per shot	Prompt each shot in a tool like Veo, Kling or Runway, then lip-sync	Rich, detailed visuals	Characters drift between shots; you pay credits for every attempt
2D puppet tools	Rig a character once and drive it with your voice or webcam	Real-time, expressive	You perform every take; rigging takes time
Animation drawn by code	An AI coding agent writes code that draws each frame of a 3D scene, and a renderer films it	Characters are identical in every episode; a one-line change re-renders in minutes	Stylised looks only, no photoreal people

The third route is how the video below was made. Nothing in it is filmed or generated by a video model: Claude Code wrote the scene, and the computer drew all 1,417 frames. It uses the same five beats as the shopping example.

Friday Deploy, 47 s. Beat 1: "Ship it to production." / "It's Friday." Beat 5: "Enjoy the hike."

If you want that exact setup, the Claude & Pixel Guide walks through it: you paste one prompt into Claude Code, it interviews you about your idea, you approve the story and pick the voices, and you get a finished 1080×1920 MP4. It needs a Mac, and it is an independent course, not made by Anthropic.

Step 5: Captions, music and sound

Captions: big, white with a dark outline or shadow, centred in the lower middle of the frame, 2 to 5 words at a time. Keep them clear of the bottom fifth, where each app puts its own buttons and text.
Music: a quiet loop under the dialogue, then drop it out completely just before the last line. The silence tells the viewer the punchline is coming.
Sound effects: one small sound on each insert (a cash register for the order, a ping for the budget alert). More than that crowds the voices.
Loudness: mix the voices well above the music. If you can hear the music over a line, it is too loud.

Step 6: Post and test

Put beat 1 on screen in the first second. The opening caption is your hook, so it should be the plan, not a title card.
Post natively on each platform rather than sharing a link to one.
Name the series and use the same name in every caption, so people can find the next episode.
Test hooks on your best episode. Make three versions with different first lines and post them a few days apart. Keep the one that holds viewers past three seconds.
Batch your production. Writing five scripts in one sitting is faster than writing one script five times, because you stay in the characters' voices.

What goes wrong, and how to fix it

Symptom	Likely cause	Fix
Viewers drop off in the first 3 seconds	The first line sets up instead of stating the plan	Open on A's plan and B's one-line fact
The joke doesn't land	Beat 3 doesn't flip the blame	Make the fix reveal that A caused the problem
It runs over a minute	Too many lines in beat 2	Keep only the lines that grow the number
It feels flat	Both voices have the same energy	Make B slower and flatter, and add pauses
Episode two feels like a different show	The characters or voices changed	Fix the look and voices once and reuse them

Frequently asked questions

How long should an animated dialogue short be?

Between 40 and 50 seconds works well. It is long enough for five beats and short enough to rewatch. Under 30 seconds, the escalation in beat 2 has no room.

Do I need to know how to animate?

No. Dialogue shorts are mostly close-ups and cuts, and each of the three routes above can produce them without traditional animation skills. You do need a clear script and consistent characters. If you'd rather have an AI do the animating in code, see can Claude make animations? and how to make videos with Claude Code.

Can I use AI voices in videos I earn money from?

Usually yes, but it depends on the voice provider's terms and your plan. Check the commercial-use terms before you monetise, and keep a note of which plan you used.

How many episodes should I make before judging the format?

Plan for at least five. A recurring cast takes a few episodes to build recognition, and five scripts give you enough to see which ideas your audience responds to. To turn those episodes into a channel, see how to start a faceless animated YouTube channel.

The course this guide comes from

The Claude & Pixel Guide: Animated dialogue videos for Reels, Shorts and TikTok, made in code from one prompt. 15 text lessons in three tracks.

Keep reading

How to start a faceless animated YouTube channel: Start a faceless animated channel with two fixed characters: the series bible, an idea bank, batch episodes, hook tests, a weekly routine and YouTube's rules.
Free AI voices and lip sync for cartoon characters: How to give a cartoon character a free voice and a moving mouth: open-source voice models, three ways to lip sync, and the licence terms to check first.
How to keep the same character in AI animation: Why AI characters change between shots and how to stop it: a character sheet, a fixed identity block, reference images, short shots, and when to use code.