claude-faceless-shorts-creator skill
Five Claude Code skills that produce faceless vertical YouTube Shorts from Remotion code, video models or paper collage
- 290 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Uses paid ElevenLabs and fal.ai keys from the `.env` file and spends money on voice and video models
- Runs Python tools, `npm` and `ffmpeg` that write files across the project
- Downloads and installs `rembg` and Chromium for collage layers
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/claude-faceless-shorts-creatorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Uses paid ElevenLabs and fal.ai keys from the `.env` file and spends money on voice and video models
- Runs Python tools, `npm` and `ffmpeg` that write files across the project
- Downloads and installs `rembg` and Chromium for collage layers
This is third-party code. Review the repository files before installing.
What it does
The repository turns Claude Code into a workshop for short vertical videos with no presenter. You describe the video in words and the skill picks one of three pipelines. `/make-short` draws every frame in Remotion code, with no footage or stock. `/make-ai-short` animates a locked recurring character through a video model on fal. `/make-vox` builds a documentary short from layers in a paper-collage manner. All tracks share a backbone. The script and the `beats.json` contract follow a six-beat grammar, from a hook at frame zero to a loop. The ElevenLabs voice returns per-word timestamps, so captions highlight the spoken word. Before voicing, Claude renders frames at phone scale and reads them itself. `/suggest-sfx` picks sound effects from a growing library, and `/vidtsx-2d-generator` holds the rules that keep Remotion renders from crashing. The folders contain 12 finished code-built shorts, one character for the generative track and a coffee collage. The `brand.md` file sets the palette, fonts and sound taste: replace it with your own and new videos follow it.
Who it is for. For short-video and educational channel creators who need a steady output of videos without a camera or an editing app.
Good fit when
- You want a channel of shorts about chess, math, algorithms or developer tips, built entirely in code
- You need a series with one recurring character generated by a video model
- You want a documentary short in a collage style with maps, archival layers and a moving camera
- Captions must highlight each word exactly when it is spoken
- You want to rebuild one of the 12 examples and swap the voice and text
Not a fit when
- You need videos with real footage or stock clips: the code track uses neither
- You have no ElevenLabs key: voice, sound effects and music go through it
- You want to build outside Claude Code: the skills live in `.claude/skills` and are made for it
- You need long-form editing: the format is built for vertical videos of about forty seconds at 1080 by 1920
Example request
Make a short about the multiply-by-11 trick in Remotion code, with an ElevenLabs voice and word-highlight captionsLimitations
You need Claude Code, Node.js 18 or newer, Python 3.10 or newer, `ffmpeg` on PATH and an ElevenLabs key for voice, effects and music. The generative track adds a fal.ai key, and collage layers need `pillow`, `rembg` and `playwright` with Chromium. Keys go into `.env` following `.env.example`.
Voice and video models are paid. The `/make-ai-short` skill must state the cost before spending money. Rendered videos are not committed because they can be reproduced from the sources. Video-model clips cannot be reproduced, so they are committed.
The titles and hooks of the examples in the repository are in English. The author promises to add links to published versions later.
How to disable. Delete the cloned folder or remove `.claude/skills` from it and Claude Code stops offering the five commands. Delete the keys from the `.env` file.
FAQ
How do I install claude-faceless-shorts-creator in Claude Code?
Clone the repository, copy `.env.example` to `.env` and fill in the keys. Then run `cd remotion && npm install && npm run gen` and start `claude` in the root. The five skills load from the `.claude/skills` folder.
How do the three tracks differ?
The TSX track is built entirely in Remotion code without footage. The generative one keeps a character and hands animation to a video model on fal. The collage track splits a scene into layers, cuts out objects and moves the camera. The wording of your request selects the track.
What are the six beats in the script?
A grammar for a forty-second short: hook (frame zero serves as the thumbnail), setup, quiz, reveal, twist and a seamless loop. The `/make-short` skill checks that the video loops back into its start.
Why does `/make-ai-short` lock the character?
A rule of the skill: a locked character is never regenerated from text. It is stored as a `character.json` sheet with a reference image and shots are built from it. Looping is achieved by an end-frame constraint.
Which examples ship in the repository?
Twelve shorts in `shorts/` from chess to physics, a generative short `The Door` with a character in `ai-shorts/blue-man`, and the `vox-1-coffee` collage. Each has a script, a beats contract and a sound-effects cue sheet.
Discussion
No questions or reviews yet.
Related
SkillsClaude Code skills that turn a talking-head recording into a YouTube video with graphics, sound and thumbnails
A skill for Claude Code and Codex that builds cinematic product videos with Remotion from shot recipe cards, a ready template and audio assets
Claude Code Video Toolkit
claude-code-video-toolkit
A video production kit for Claude Code: script, voiceover, music, visuals and MP4 rendering through Remotion, with some steps on paid cloud services
A Claude Code skill that turns an idea into a five-stage prompt for Sora, Kling, Veo and Seedance, with templates for 21 genres
Two Claude Code skills for videos made entirely in code: hand-painted watercolour animation and kinetic typography reels
A Claude Code skill that turns one sentence into a vertical AI micro-drama, from script and storyboard to video
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| claude-faceless-shorts-creator (this entry) | no | 290 | |
| anything2explainer | no | 2,835 | makes a narrated Remotion explainer from a topic, without video-model or collage tracks |
| Claude Motion Design | no | 725 | draws motion videos in HTML and ffmpeg without an ElevenLabs voice or word-level captions |
README in short
The README presents a table of phrases and tracks, then the list of examples and a flow from topic to the final file with effects. It continues with the five skills, a quick start, the folder layout and a link to the author's guide.
README badge
Original SKILL.md
--- name: make-short description: Build a fully-synthetic vertical short (1080×1920, ~40s) end-to-end from a topic — script + beats.json, linked TSX beats over a persistent canvas, phone-scale QA, render, ElevenLabs voice with exact word-synced captions, SFX audition, optional music bed. Use when the user wants to "make a short", "create a shorts video", build a TSX short for a niche (chess, math, algorithms, dev tips, Excel…), add voice/captions to a short, or continue the short-N series in this repo. This is the 100%-TSX track — for video-model pixels use make-ai-short, for the paper-collage documentary look use make-vox. Defers raw TSX crash rules to vidtsx-2d-generator and SFX taste to suggest-sfx + brand §7. --- # make-short — fully-TSX vertical shorts, end to end Everything is code: no footage, no stock. One short = one Remotion composition fed by a beats contract, with voice and SFX layered on after render. Proven across the 12 example shorts in `shorts/`. The idea bank + niche ranking lives in **`IDEAS.md`** — read it when picking a topic, and grow it when brainstorming. Run everything from the **repo root**. Tools are stdlib-only Python 3.10+; `ffmpeg`/`node` must be on PATH; API keys live in `.env` (see `.env.example`). ## Artifact contract ``` shorts/short-N-<niche>/ script.md — hook, beat sheet table (time | on screen | VO), production notes beats.json — machine contract: format, vo[] (text+start/end+words), beats[], niche timeline voice/ — per-line TTS clips + voice.wav (gen_voice.py) [gitignored] sfx-plan.json — cue sheet on the global timeline (mix_sfx.py) output/ — short-N-sfx.mp4 (+ music auditions) [gitignored] remotion/src/shots/short-N/ ShortN<Name>.tsx — THE composition (registered by npm run gen) vo.gen.ts — AUTO-GENERATED VO with exact word times (never hand-edit) ``` Shared kit: `remotion/src/lib/shorts.tsx` (Captions, Kicker, BigTitle, Stamp, PauseCard, StatChip, ProgressBar, ShortsBackdrop, SAFE areas, prog helper). Niche libs so far: `lib/chess.tsx`, `lib/math.tsx`, `lib/algo.tsx`, `lib/sheet.tsx`, `lib/piano.tsx`, `lib/prob.tsx`, `lib/map.tsx`, `lib/orbit.tsx`, `lib/chart.tsx`, `lib/story.tsx`… **Add at most ONE new niche lib per short**, generic enough for a series. ## Stage 1 — script + beats Beat grammar (~38–42s): **HOOK** (0–3.4s, frame 0 FULLY composed — the payoff already visible, no fade from black) → **SETUP** → optional **QUIZ** (PauseCard, ~2.5s — drives comments) → **REVEAL** in 2–4 steps, each synced to a VO word → **TWIST** → **LOOP** (last frame == frame 0, dissolving the payoff back into the intro so it replays seamlessly). **Outros — no fluff.** NEVER end on an old-school engagement-CTA — no "what should I do next?", no "comment below", no "which one should I break down". They read as dated. End on the PAYOFF line, and let the visual LOOP blend back into the intro — the loop-into-intro *is* the ending. Prefer a clean visual dissolve/reset that lands the last frame exactly on frame 0. If a seamless loop genuinely isn't possible for a topic, just end on the payoff with no filler. Don't pad the tail: trim the composition so it ends shortly after the payoff, not seconds of dead air. VO rules: ~100 words total, estimate line windows at **~2.7 words/sec with slack to the next