under the hood (Claude Code skill)Partly tested🌏 World

Rajesh made a Claude Code skill that turns a tech topic into a 1080p explainer with hand-drawn animation, a cloned voice and sound effects, all on your machine

Rajesh (@royalreplaced) open-sourced under the hood: in Claude Code you say "make a video about what happens when you git push", and the skill checks your setup, asks one or two questions (which angle, which voice), shows you the narration and storyboard to approve, then writes a canvas animation, screenshots every scene and fixes overlaps itself, generates two takes per line locally with the open-source TTS model Chatterbox for you to pick, and renders a 1080p MP4 with music that ducks under the voice, a sound on every beat, and loudness at -16 LUFS for X and YouTube. You can also record 20–30 seconds of your own voice to clone. His demo is "what happens when you type cookie into Google". On Linux we rendered the repo's bundled cookie example without narration: a 33.4-second 1080p30 video in about 27 seconds, with complete frames. We did not test narration, voice cloning or Claude writing a new topic.

Visit site under the hood (Claude Code skill) ↗
Pricing
Free
Region
Global
Where it's from
Overseas
Added
Nov 13, 2026
Tested on
Nov 13, 2026
Version we used
Rajesh-Royal/under-the-hood git c145509; playwright-core 1.56.1; bundled cookie example, --silent
Categories
VideoCoding
Original author
Rajesh @royalreplaced ↗
Our test
Partly tested
Cost / time
The skill is MIT and free; voice and rendering run locally with no per-video fees. Writing the storyboard and animation uses your Claude Code plan.
Source
X @royalreplaced ↗GitHub Rajesh-Royal/under-the-hood ↗

The task

You want short "what actually happens when…" tech explainers but can't animate, and don't want to pay for voiceover or cloud video generation.

Steps

  1. Install in Claude Code: /plugin marketplace add Rajesh-Royal/under-the-hood, then /plugin install under-the-hood@under-the-hood. Or copy skills/under-the-hood and skills/under-the-hood-setup from the repo into ~/.claude/skills.
  2. Name a topic in plain words, like the prompt below, or use /under-the-hood:under-the-hood what happens when you git push. The first run checks for and installs ffmpeg, Node 18+, uv and playwright, asking before every brew or apt command.
  3. Answer one or two questions: pick an angle for a broad topic, and a voice (built-in, saved, your own from a 20–30 second phone recording, or none).
  4. Approve the narration and storyboard table. Changing words here is free; after voice generation it costs a re-run.
  5. It writes scenes.js, screenshots each scene and fixes layout; then it generates two takes per line, which you judge by ear (it can't hear audio) and pick line by line.
  6. It renders the MP4; open the project's index.html for a preview with chapters. Editing one narration line re-syncs the whole video.

Prompt

make an under-the-hood video about how HTTPS works

What the author says

Rajesh @royalreplaced (X, 12 Nov 2026): hand-drawn animation, your own cloned voice, music and sound effects, 1080p, all running locally. README: a 60-second video renders in about 30 seconds on an M-series Mac; frames render deterministically in headless Chrome; music and effects are synthesized with Web Audio, so no sample licensing; the voice model needs about 2 GB of disk. Tested on Apple Silicon macOS; Linux and WSL "should work".

Our test

On 13 Nov 2026 on Linux we cloned Rajesh-Royal/under-the-hood (c145509, MIT), installed playwright-core 1.56.1 (the version pinned in the setup script) into a temp folder and pointed CHROME_PATH at the local Chrome. We created a project with new_project.sh --example cookie and ran node render.js cookie-test --silent (no audio): 33.40 s, 1002 frames, rendered in about 27 s to a 1920×1080, 30 fps, ~4.1 MB MP4. A frame at 20 s showed step 5, "Thousands of machines search the index at once": the inverted-index card, lines from the cookie query to 24 servers, the top progress tracker and bottom caption all intact. We did not install Chatterbox, generate narration, music or the sound mix, or have Claude write a new topic.

What we found when we tried it:Frame rendering works on Linux as-is, at roughly the speed the author quotes (about 30 s per 60 s); the script and engine are solid, so quality mostly depends on Claude's storyboard and the voice.

How to get in

The original post and README are in English; the project page https://rajesh-royal.github.io/under-the-hood/ has the demo video.

Who it suits

Developers, tech creators and teachers making short explainers. The skill calls Chatterbox's Turbo and standard models and its narration rules are written for English; we have not verified Chinese narration.

Pitfalls

  • The voice model downloads about 1 GB of weights on first use, and generation is slower than real time without a GPU.
  • Claude can't hear audio, so you listen to and pick every take yourself.
  • Only clone your own voice or one you have permission for; Chatterbox adds an inaudible watermark.
  • Avoid em dashes and semicolons in narration and on-screen text; the TTS reads them oddly (the skill enforces this).
  • We only rendered the bundled example without narration on Linux; we did not test voices, cloning or Claude writing a new topic.

More in this category