CryptoYuyyy02/stickman-video-director
Media & DesignA Muse agent skill that walks through scene generation, voiceover, BGM and editing to produce full stickman story videos.
This repo is an agent skill pack · 1 skillsSkill list and install →What it is
This repository packages a production workflow for making stickman story shorts with the Muse agent. Installing the skill lets Muse break a story into scenes, generate footage, record voiceover, synthesize background music and edit the result into one video. The README documents a 58-second landscape sample, The Elevator Stops at Empty Floors, produced with a six-scene split, Chinese voiceover, code-generated BGM and ffmpeg assembly with subtitles and loudness normalization (README).
The skill is a set of instructions and reference documents, not a standalone program. It adapts the MIT-licensed upstream project by kaomei and adds a Muse-specific pipeline and revision notes (upstream).
How it works
The method centers on a few rules:
- Two engines: a model engine that generates each scene directly for polished output, or a code engine that animates frame by frame with cheaper, more controllable results for drafts.
- Four-part prompts: reuse a fixed character and art style verbatim across scenes, changing only scene and timed action so the character does not drift.
- Scene splitting: keep single generation under roughly 10 seconds and split longer films by beat before reassembly.
- Timing lock: voiceover starts at 0 seconds, scenes are speed-adjusted to the narration length, and segments are joined with no silence.
- Review on real footage: sample frames per scene to check character consistency, beat order and contact relationships, and redo failures.
Quick start
-
Have a working Muse setup with a custom skills directory.
-
Copy the whole repository folder into your skills directory, for example:
~/.claude/skills/stickman-video-director/Or use the equivalent custom skills directory of your agent framework.
-
Ask Muse to make a stickman video. It should follow the rules in
SKILL.mdautomatically.
SKILL.md is the main skill file; UPSTREAM-SKILL.md holds the complete upstream text; references/ contains the A7 prompt contract, acceptance rules, storyboard templates, style catalog and an identity reference image (README). The README supplies no sample prompt or expected output for a first run, so expect to iterate against generated footage.
Constraints
- Requires Muse or a compatible agent framework that loads custom skills from a directory; it is not run on its own.
- The workflow assumes access to video-generation models and ffmpeg for assembly, subtitles and loudness normalization.
- The MIT license carries over from the upstream project, but this fork's license file is marked NOASSERTION by GitHub; check
LICENSEbefore redistribution. - Quality depends on the underlying video model and on following the fixed-character prompt discipline; the README's sample is one documented production, not a guarantee for other stories.
HotRepo's take
The repository appeared on HotRepo's new-project board on 2026-11-21, gaining 10 stars in about an hour and ending at 113 stars and 12 forks. It is a prompt-and-workflow skill rather than code, so its value rests on the documented production rules and reference files; readers weighing it should compare those rules and the sample film against the upstream project to see what the Muse adaptation actually adds.
Written by AI from the project's public information and HotRepo's records, updated . The project's own docs are the source of truth.
You might also like
- win4r/video-directorA Claude skill that guides coding agents through planning, building and QA-checking code-rendered short videos.
- konlu/story-studioAn agent skill that turns a topic into a scripted, voiced, illustrated short video and an editable JianYing draft.
- threerocks/sweety-handdraw-videoTurns a topic, article or story into a narrated hand-drawn explainer video, with stroke-by-stroke drawing and voiceover timed to the script.
- andyhuo520/digital-human-kouboTurns one photo and a script into a packaged talking-head video, driven by an AI agent instead of manual editing.