Creator OS
ai videoclaude skillclaymationshort form videotiktok

Claymation Claude Skill: Make Stop-Motion Shorts in 2026

A Claymation Claude Skill turns a prompt into a stop-motion-style short: frames, timing, captions. See how to set it up in Claude Code and ship it to TikTok.

Creator OS · November 14, 2026 · Updated November 16, 2026 · 9 min read

What Is a Claude Skill for Stop-Motion Video?

A Claude skill for stop-motion video is a set of instructions that tells Claude Code how to plan and render a stop-motion style short: frame by frame timing, captions, hooks and a rendered video file on your own machine. You type a prompt. The skill handles the shot list, the frames and the edit, then you publish the result.

The short answer for search: the skill does not generate stop-motion footage with a video model. It builds the look from still frames and a local render, using Python and ffmpeg on your computer, in the same way the open source OPEN VIDEO EDIT skills turn a talking head clip into an animated short. The skill is the recipe. Claude Code is the cook.

your prompt
   |
   v
Claude Code + your stop-motion skill
   |
   +-- shot list (scenes, timing, hook, caption plan)
   +-- frames (images or generated stills per beat)
   +-- local render (Python + ffmpeg)
   |
   v
mp4 short  -->  creatoros posts:create  -->  TikTok / Shorts / Reels
From prompt to published stop-motion short.

Get started for $0

How Does a Stop-Motion Style Skill Fit With Other Claude Video Skills?

It sits next to the other Claude video skills as one look among several, not a replacement for any of them. Each skill is a style and a pipeline. The shared pieces are Claude Code, local rendering, and a publish path.

Skill family Input What you get
Stop-motion style A prompt or a short script Frame by frame short with captions and a hook
Animated talking head A raw talking head clip Same clip with captions, logos, b roll cards, counters
Fireship style edit Screen recording or clip Fast cut edit with overlays
AI launch video One line of text A launch style video built from that line

The talking head and Fireship skills ship in the open source OPEN VIDEO EDIT repo. Its skill list is split-animated-talking-head, longform-animated-talking-head, fireship-style-edit, fireship-style-short, split-animated-short and reverse-engineer. A stop-motion style skill you write yourself follows the same pattern: a name, a prompt contract, and a local render step. For the base install, the one line skill walkthrough at the AI launch video generator post shows the format.

What You Need Before You Write the Skill (Claude Code setup)

You need Claude Code, Python, ffmpeg and a folder to render into. Nothing else is required to make a video that sits on your disk.

  • Claude Code installed and signed in. Skills are files Claude Code reads from your project.
  • Python and ffmpeg on your PATH. Renders happen locally.
  • A model. Any of Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 4.5 or Claude Fable 5.1 works with the Creator OS side. For heavy frame planning, a larger model pays for itself.
  • Frame assets if you want a photographic feel: images from your own library or generated stills, dropped into an assets folder the skill can read.
  • A publishing path. Creator OS accounts if you want the skill to post the finished file without you opening each app.

Before you render anything, test the caption rules. A stop-motion short has short beats, so captions must fit. Run:

creatoros validate:post-length --text "Clay figure tries to
open a jar for 40 seconds"

That check is one network call and it saves a re-render. If you plan to let the agent publish, install the Creator OS side once:

npx @creatoros/cli@latest init

That command installs 14 Claude Code skills, including post-shortform and post-everywhere. Details are in the Claude Code plugin docs. If you want the plugin route instead, it adds the MCP server, the campaign builder skills and a mod with a live pane and a publish guard that asks before anything is published.

How to Write and Run the Skill Step by Step

The flow is: add the skill files, add a brand file, run one prompt, review the draft render, then publish from the same session.

  1. Add the skill. Drop your stop-motion skill folder into your project skills directory. Keep the skill name short, for example stop-motion-short.
  2. Write a brand file. One page: voice, colors, logo path, fonts, what you never say. The skill reads it on every run, so you stop repeating yourself.
  3. Prompt it. Give the topic, the length and the platform. Say the beat count. Vague prompts produce vague timing.
  4. Review the draft. Open the rendered mp4 before anything publishes. Fix the hook line first, since that is what decides retention.
  5. Publish. Either upload the file by hand or let the agent post it. The CLI post command takes a media id or a URL.
creatoros media:upload ./clay/render-final.mp4
creatoros posts:create \
  --text "Day 3: the clay intern negotiates with a stapler" \
  --platforms tiktok,youtube,instagram \
  --media med_abc123 \
  --scheduledAt 2026-06-02T17:00:00Z --timezone America/Toronto

Or through the MCP tool, from inside the same Claude Code session:

create_post
  content: "Day 3: the clay intern negotiates with a stapler"
  platforms: ["tiktok","youtube","instagram"]
  media: ["med_abc123"]
  schedule_at: "2026-06-02T17:00:00Z"
  timezone: "America/Toronto"

Creator OS applies each platform’s own rules on the way out: Reels handling, Shorts detection for vertical video, TikTok consent and privacy flags, and a draft flag if you want it to land in the TikTok inbox instead of going live. Watch the walkthroughs on the KevBuildsApps YouTube channel if you prefer to see it done before you run it.

Can a Stop-Motion Style Skill Make Launch Videos and Faceless Shorts?

Yes to both, if you keep the format tight. A stop-motion style launch video works because the look is odd enough to stop a scroll, and a faceless short works because the skill never needs your face: the characters are the cast.

For a launch video, structure it as five beats. Problem, product, one demo moment, proof, call to action. Stop-motion gives you a visual metaphor for each beat, which is easier to write than a live action script. A founder who does not want to be on camera can build the whole launch this way.

For faceless shorts, pick a recurring character and reuse it. A small clay figure with one job, one problem and one outcome per short. Serial characters compound. One-off shorts do not. Keep a character sheet in the brand file so the skill renders the same figure every time, and keep a numbering convention in the output folder so you can find the last episode fast.

When the topic touches Claude Code, local rendering or the open source tools, the launch video at youtube.com/watch?v=-QBJH_PK3pY is the best single reference for how the pipeline behaves end to end.

What the Skill Outputs: Frames, Timing, Captions, and Hooks

Treat the output as four files, not one, so you can fix a caption without re-rendering the whole short.

Output What it holds Why you care
Frame plan Scene list with frame counts and per beat timing Change pacing in seconds, not hours
Assets folder Every still used in the render Swap one image without a rebuild
Caption file Burned in text plus a sidecar subtitle file Silent autoplay still reads
Final mp4 Vertical render at your target length Ready to upload or publish from the agent

The hook is a separate line in the plan. Write three, render the strongest, keep the other two for the follow up. Stop-motion reads slower than talking head video, so a hook that needs four seconds to land is too slow. Two seconds is the target.

After publishing, measure. Creator OS returns per post analytics and Reel watch stats through get_post_analytics, including reelsSkipRate, the share of views that swiped away in the first 3 seconds, plus igReelsAvgWatchTime in milliseconds. An agent can read that number and adjust the next hook. The numbers come from Creator OS. What to change is your call or your prompt’s call. The analytics side is covered in the agents docs.

Where to Post: TikTok, YouTube Shorts, and Social Feeds

One vertical render covers all three, as long as you respect each platform’s own rules on the way out. The file is the same. The settings are not.

Platform Setting that matters for a stop-motion short
TikTok privacy_level, allow_comment, allow_duet, allow_stitch, plus a custom cover image
YouTube Vertical short, title up to 100 chars, firstComment posted and pinned, no custom thumbnail on Shorts
Instagram Reel with a custom cover, plus a firstComment for any links, since captions are not clickable

Post the same file to all three in one call and let Creator OS handle the flags. Networks you name but have not connected come back in missing_platforms, and the rest still go out. That means a typo in the platform list does not kill the whole post. Full option reference lives at creatoros.ca/docs/post-options.

If you want a text based companion piece, the SEO blog use case at creatoros.ca/docs/use-cases/seo-blog turns the same topic into a WordPress article. One topic, one short, one article.

FAQ

What exactly does a Claude skill for stop-motion video generate?

It generates a plan and a rendered video file: a scene list, a set of still frames, a caption file and a vertical mp4 assembled locally with Python and ffmpeg. It does not call a video model to invent motion. You review the file, then publish it, either by hand or from the same Claude Code session through the Creator OS MCP tools.

How is this different from the Claude launch video skill?

The launch video skill starts from one line of text and produces a launch style video. A stop-motion skill is a look: the output is a stop-motion style short, and you usually supply a script or a shot list. You can combine them, using the launch skill’s script as the input to the render, as long as the timing and beat count match your target length.

Can I use it for faceless shorts?

Yes. A stop-motion short has no presenter by design, so there is no face to show and no camera to set up. Reuse one character across episodes, keep a character sheet in your brand file, and number the output files so you can track a series. Serial characters build recognition faster than one-off shorts.

Do I need the post-shortform skill too, or does this replace it?

They do different jobs. Your render skill makes the video. Publishing, platform flags, scheduling and replies belong to Creator OS skills like post-shortform, plus the MCP server and the CLI. You want both: one to render, one to get it out and answer the comments that follow.

How much does it cost to run?

The skill itself is files in your project, so there is no separate skill fee. You pay for the model calls you use while planning and rendering, on whatever Claude plan or API key you already run. On the Creator OS side everything is included from $14.99/month, with the API, the MCP server and the CLI in every plan.

Get Started

Render the short locally, then let the agent publish it. Start at creatoros.ca/sign-up. Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included. New subscribers start for $0 with a free trial and can cancel anytime. The full setup reference is at creatoros.ca/docs.

Get started for $0
Free trial for new subscribers. Cancel anytime.

Keep reading