Make an Explainer Video

1. Commit to a topic in computer science or math, and decide what a viewer should understand after a minute of your video. Write your topic, that sentence and your stack in the text box of the Week 6 quiz’s last question, before the lab. This is the topic your video explains in the lab. Don’t research it in depth: you work out the details with your agent in the lab.
2. Set up your stack on your own laptop, and render a five-second test clip with one narrated sentence. Your lab instructor checks it at the start of the lab.

A recommended stack

• Agent: Claude Code with Opus 5.5 or higher (/model), or Codex with gpt-6-astra (gpt-6.1-sol if you reach your plan’s limits).
• Animation: Remotion writes the video in React and renders an MP4; its Studio shows the video live in the browser while your agent edits it. It needs Node.js. Add its skills for your agent with npx skills add remotion-dev/skills. On Linux or WSL, first install the libraries its browser needs. Other stacks are welcome too, for example an HTML page with an SVG or JavaScript animation. Choose one that previews quickly.
• Narration: Kokoro is free text to speech that runs on your laptop’s processor. Install kokoro-onnx and its model files (about 120 MB). It speaks English and a few other languages, not Hebrew or Arabic.
• ffmpeg puts narration under the video and compresses it: brew install ffmpeg on a Mac, sudo apt install ffmpeg on Linux or WSL.

Start two things at once, in two windows. In one, have your agent put a first visual on screen right away and render it, to test your stack and the rendering. In the other, write a more complete script with your agent: what each scene shows and what the narration says. Keep it in SCRIPT.md.

Keep every step small: one scene, or one change to a scene. A step takes a few minutes, not half an hour. If an agent is building a lot at once, stop it and split the work. Look at the preview rather than rendering the whole video.

Work in parallel. Keep each scene in its own file, so two agents can work at once: while one builds the next scene, review the last one and have another agent fix what you find. Commit after each step, with a message saying what you changed and why. Render the whole video only near the end. A partial video is fine.

Draw illustrations in code; screenshots are welcome when useful. Do not use images made by generative models.

1. Collect. Commit your work. Download the collect skill, SKILL.md, into .claude/skills/lab06-collect/ without committing it. In a fresh agent session, run /lab06-collect. In Codex, ask the agent to read the file and follow it. It compresses your latest render into lab06-video.mp4 and packs your source, SCRIPT.md and a log of your iterations into lab06-bundle.zip.
2. Share it for the play. Drop a copy of lab06-video.mp4, named after you (for example your-name.mp4), into your section’s shared folder. Your lab instructor gives you the link in the room.
3. Upload both files to Lab 6 report on Moodle before the play starts. This upload is what is graded.
4. Watch. Your lab instructor plays the start of everyone’s video from the folder.

Finished your video after the lab, and proud of it? Run the collect skill again and upload its two files to Lab 6 finished video on Moodle, open until December 3. It is optional and not graded. By uploading there, you agree that we may show your video in a gallery on the course site.

Lab prep question in the Week 6 quiz: 1 point. Setup working at the start of the lab: 1 point. Video and bundle uploaded before the play: 2 points. A partial video counts. Max grade: 4 of 4 points.

Late for the lab (over 10 minutes): -1 point

Not graded: your topic, your stack, or how polished or complete your video is.