5 Styles to Turn Ideas into Stickman Videos Stickman Video Director

Explain one problem through a stickman story

You may have a topic but no clear idea of what should happen on screen. Stickman Video Director turns source material into narration, character actions, and a storyboard. Review the story first, then get prompts for the individual clips. The skill handles planning; a video tool such as Flow generates the clips, and an editor such as CapCut assembles them and adds subtitles.

Compared with 16 Styles to Turn Ideas into Explainer Animations, which uses code to make style previews and delivers an editable HTML animation, this skill produces an approved proposal followed by prompts for a video model. It also differs from the hand-drawn whiteboard workflow: that workflow synchronizes drawing with narration, while this one advances a story through character actions, props, and camera changes.

The three official clips on this page show the modern tech, dark minimalist, and cozy styles, in that order. Each lasts about ten seconds. Watch how the character moves the scene forward and how the props and setting support one idea. These are style samples, not a complete sixty-second film.

For this tutorial, make a thirty-second film about choosing a location for a coffee shop. The point is simple: many people walking past does not necessarily mean many people will enter.

Choose a style and prepare your first video

Install the skill and confirm it is available

1. Send this installation request in Codex chat:

   Install this skill:
   https://github.com/kaomei/stickman-video-director
   Confirm that directing-stickman-videos is available and list its visual styles.

2. Confirm that Codex can find directing-stickman-videos and that its installation includes references/. Those files contain the storyboard, style, and voice rules; copying only SKILL.md leaves out part of the skill.

The skill itself needs no API or MCP configuration. You will also need access to a video generator. This tutorial follows the creator’s route: Flow for generation and CapCut for editing.

Provide the source and production choices together

The current repository has five style families and seven visual options. The earlier post does not describe every option now available:

Style	Options	Useful for
1: Classic Minimalist	Black lines on white, or white lines on black	A concept expressed through characters and a few graphics
2: Modern Beanie	2A Modern Studio Tech, 2B Cinematic Story	2A uses bright spaces and floating interfaces for tech explainers; 2B uses cinematic lighting and depth to set the mood of a story
3: Prehistoric Hand-Drawn	Dark comedy	Satirical stories told through prehistoric characters and props
4: Cozy Storytelling	Full-color 2D settings	Everyday experiences, emotions, and character stories
5: Historical Cartoon	Era-specific clothing and settings	History explainers and stories about historical daily life

Choose Cozy Storytelling for this example so viewers can follow the shop owner’s experience. Send the following request in Codex. It is an editorial teaching example based on the skill’s rules, not the prompt behind an official sample; we have not generated a film from it.

$directing-stickman-videos
Make a 30-second, 9:16 stickman video in Style 4, Cozy Storytelling.
Audience: people preparing to open their first coffee shop.
Question: why might a busy street fail to bring customers into a shop?
Source: a shop owner considers a unit near a subway exit, assuming a busy location is enough.
They observe that most morning commuters hurry past. At another corner, fewer people pass,
but some wait for friends, linger, or enter nearby shops.
Use this situation to explain why location research should consider the intended customers
and their reasons to stop and buy, rather than just counting passersby.
This is an illustrative story. Do not invent footfall, conversion, or revenue figures.
Use the skill's default English narration and include a Chinese reference translation.
Prepare a director's proposal for 3 clips with narration, actions, location changes,
camera choices, and a voice profile. Wait for my approval.

Thirty seconds means three clips of approximately ten seconds. If you omit duration, the default is sixty seconds and six clips. Plan in ten-second multiples and choose either 16:9 or 9:16. If the aspect ratio or style is missing, the skill should ask before planning.

Review the proposal before requesting clip prompts

Check that the story answers one question

Read the proposal for three things: a clear opening question, a change the character experiences, and an ending that explains the cause. Attractive scenes alone are not enough.

Use the following outline to judge the proposal; Codex should supply the actual storyboard:

Clip	What the scene should do	What viewers should understand
1	The owner watches the subway crowd and imagines a coffee shop	They treat a busy street as the main reason to choose a location
2	Cut between commuters hurrying past and people lingering at another corner	Passing through and stopping are different behaviors
3	The owner records customer activities and compares the locations again	Look at who might visit and why they would stop, not just the size of the crowd

Cozy Storytelling normally changes location between clips to advance the story. If a return to the same shop matters, plan it explicitly in the proposal. Character design, important props, and the narrator should remain consistent.

Settle the narration and voice, then approve

1. Check the narration against the source and remove unsupported conclusions. The skill targets about 20–25 English words per ten-second clip; that is not a Chinese narration-length rule.

2. Set a voice profile for the whole film in the proposal, including timbre, accent, pace, and delivery. Generate and listen to the first clip before making the rest.

3. If clip two is too vague, send a concrete revision in Codex:

   In clip two, first show commuters hurrying past the unit near the subway,
   then cut to people waiting for friends at the other corner.
   Express the difference through actions, without adding numerical claims.
   Keep the owner's design, overall style, and narrator voice.
   Revise the affected narration and storyboard, then wait for my approval.

4. When the revised proposal is ready, send:

   I approve the current director's proposal. Produce 3 standalone clip prompts,
   approximately 10 seconds each, with complete character, style, and voice descriptions.
   Include notes on transitions, editing, and subtitles to add in post-production.

You should receive three prompts you can submit separately. The repository calls the director’s proposal Phase A and the clip prompts produced after approval Phase B. Next, paste each complete prompt into Flow as text to generate its clip. Each prompt should contain the descriptions needed for that clip, rather than relying on a bare reference such as “same character as above.”

Generate the clips and assemble the film

1. Submit the first prompt in Flow with the approved aspect ratio. Play the result with sound. Check for complete narration, missing words, and actions that communicate the point. If the platform offers a different clip length, return to Codex to revise the segmentation and narration before generating the remaining clips.
2. Once the first clip works, generate the others in order. Compare each with the first for character design and voice. Check that important props do not disappear, change hands, or change appearance without a reason.
3. For a faulty clip, tell Codex the time and specific problem, then regenerate using its revised prompt. “The hat changes color at six seconds in clip two, and the narrator sounds different” is more useful than “make it better.”
4. Import the clips into CapCut in order. Play them continuously and adjust cuts and volume. The default prompts keep visible text out of the generated images and treat narration as audio only. Add subtitles in the editor from the actual audio rather than asking the video model to draw them.

Viewers should hear a clear point and see the character’s actions support it. If the narration changes, regenerate the affected audio or clip and update the subtitle timing and shot lengths.

Adapt it to your topic or product

Change the content first, then decide whether the style also needs to change. For a booking tool, the owner might move from checking scattered messages and writing appointments down to confirming bookings in one place. Show functions your product actually has; do not invent productivity figures.

Send this template in Codex after replacing the brackets with your own information:

$directing-stickman-videos
Explain [one problem] to [audience], so they understand [one conclusion].
Source: [original text, facts, or existing product features].
Main character: [who]. Difficulty: [a specific situation].
Change to show: [actions and outcome].
Duration: [a multiple of 10 seconds]. Aspect ratio: [16:9 or 9:16].
Style: [number and option].
Use English narration with a Chinese reference translation.
Prepare the director's proposal first and wait for approval before writing clip prompts.
Do not add unsupported figures or claims. Use one narrator voice throughout.

For a single action change, check the transitions into and out of that clip. For a new style, ratio, duration, narration, or overall structure, request and approve a revised director’s proposal before generating new clip prompts. Switching from landscape to portrait requires new character placement and movement, not just a crop.

See the skill rules for the full style and prompt specifications, and the original long-form article for the creator’s topic and production workflow. The repository uses the MIT license.