Oct 27, 2026
Make Faceless YouTube Videos With the Stickman Video Director Skill
Stickman Video Director turns a topic, article, or link into a stick-figure video with consistent art direction: Director Mode gives you storyboards and scene-by-scene prompts to generate and edit yourself, while Autopilot Mode runs inside Codex or Claude Code and delivers a narrated, captioned video file.
Running a faceless YouTube channel usually stalls at the same step: every video needs visuals that look like the last one. Stickman Video Director, a 3k-star MIT project, attacks that problem with six named stick-figure styles and two production modes — one that keeps you in control of every shot, one that hands the agent the assembly work.
Why This Skill Matters
The skill (installed from the repo's skills/directing-stickman-videos directory) covers the whole pipeline: visuals, voiceover, captions, sound effects, and subtle animation. The README's banner counts 6 styles and 8 visual options across content genres: Classic Black & White for explainers, Modern Pop for product and tech stories, Prehistoric Cave for satire and fables, Cozy Storytelling for relatable everyday moments, Historical Cartoon for history shorts, and Minimal Hand-Drawn for long-form narration.
The two modes split by who assembles the video:
| A · Director Mode | B · Autopilot Mode | |
|---|---|---|
| Approach | You control production: the agent plans storyboards and scene-by-scene prompts; you generate clips and edit them yourself | Give a topic or brief, get a finished video |
| Length | Short-first, typically 30 seconds–3 minutes | Long-form-first, especially 10 minutes or more |
| Tools | A Skill-compatible assistant, then Google Flow and CapCut | Codex, Claude Code, or another agent that reads and writes files and executes commands |
Installation
Send the README's install message to any AI assistant that supports Skills:
Install this Skill for me:
https://github.com/kaomei/stickman-video-director
Once installed, read the installed SKILL.md right away.
Explain what each mode produces, the tools it needs, and the workflow
in plain language, with examples I can copy.
Based on what I want to make, tell me what is missing and exactly what to do next.
Real Workflow: A 5–10 Minute Narrated Video
Autopilot Mode needs an environment that can read and write files and execute commands — Codex or Claude Code both qualify. After installing, start from the README's own prompt:
Use Autopilot Mode / mode B to turn the following topic, article, post, news item, or brief
into a compelling YouTube video:
[Paste content or a link, or describe what you want to promote]
Requirements: 5–10 minutes, landscape 16:9, B1 "Minimal Hand-Drawn" style, with voiceover, captions, and sound effects.
Check the image generation, voiceover, and video-rendering setup first. Then give me a script and a 20-second preview with sound.
After I approve it, build the full video with visuals timed to the narration. Deliver a complete video file; I do not want to assemble scenes myself.
You approve the script, then the 20-second preview with sound, and the agent renders the full video timed to the narration. If one visual or voiceover section comes back wrong, revise just that section — the already-approved voiceover and assets carry over.
Director Mode is the budget-conscious path: the agent produces a director's treatment, storyboards, and per-scene video prompts; you generate clips in Google Flow or a similar service and finish the edit in CapCut or your preferred editor.
Tips
- Autopilot Mode needs working image and voice generation services plus a local video-rendering environment — the README is explicit that a standard chat window cannot replace that setup.
- Director Mode exists to control cost: the README notes a full generation pass for a video over ten minutes can cost tens of US dollars, so reviewing scene by scene stops you paying for rejects.
- For consistent characters, ask the agent to design the stick figure first and approve the front, side, and back views; if your video tool supports reference images, attach the approved character image to every clip you generate.
- A and B are separate style series with their own numbering and galleries; the README says you can ask your assistant to combine a preferred A-series style with the B-series workflow as a DIY adaptation.
- More styles are planned: the README lists illustrated storybooks, doodle fights, two-character sitcoms, card-based adventures, live-action/animation hybrids, and classroom explainers.
When Not to Use This
The skill itself is free and open source, but the connected image, voice, and video services charge separately — the README cautions that a consumer subscription does not necessarily include API credits. If you want everything inside one chat window, neither mode fits: Director Mode assumes a video generator and an editor, and Autopilot Mode assumes an agent runtime plus render dependencies. Commercial use is permitted under the MIT license, while third-party images, music, and voice services keep their own license terms, and YouTube monetization stays subject to the platform's review.
See the leaderboard for more skills.