Communitygithub.com

explainer-video

Create a vertical explainer, review, or launch short video on a topic. Verify facts against official sources, generate showcase images, write a Chinese voiceover with word timings (Doubao TTS), then render a keynote-style 1080×1920 video beat-synced and pop-in timed to the spoken words, with original score and SFX synthesized in code.

What is explainer-video?

This skill empowers an AI agent to autonomously produce a 60–90 second vertical (1080×1920) explainer, review, or launch video in a keynote-style aesthetic—black background, oversized typography, full-bleed images, all cut to the beat with an original synthesized score and sound effects. It demands rigorous fact-checking against official sources, honest visual critique, and zero on-screen references to APIs, vendors, or costs. The workflow follows a strict pipeline: fact-gathering into RESEARCH.md, generating showcase images via a dedicated script, authoring a script.json (approximately 6 characters per second), performing TTS through Doubao to produce word-timed audio, and then rendering the final video using Python (PIL + numpy) and ffmpeg. The skill is designed for platforms like Claude Code, Cursor, and Codex, and is particularly useful for developers who need a fast, code-driven video production pipeline without any video editing software or stock media. The provided worked example (`examples/gpt-image-2.6/`) serves as a template; users are instructed to copy and adapt rather than start from scratch.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/rayest/ai-video-skills/tree/HEAD/skills/explainer-video

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Topic explainer short video

A 60–90 s vertical video (1080×1920) in a "keynote" look: black background, very large type, full-bleed images, beat-synced cuts, an original score and sound effects, all rendered with Python (PIL + numpy) and ffmpeg. No video editor, no stock music.

Worked example: examples/gpt-image-2.6/ (RESEARCH.md, script.json, render.py, audio.py, build_all.sh) produced a 92 s, ~20 MB video about an image-model launch. Copy and adapt it rather than starting from zero. Its showcase images are not in the repo; generate your own with scripts/gen_image.py. Read references/render-pipeline.md before any render.

Settings come from the environment or a .env file (see .env.example). Never print keys, their lengths or prefixes.

Content rules (learned the hard way)

  • Result in the first 2 seconds. Open on the most impressive output, not on a title.
  • Nothing backstage on screen or in the VO: no API relay or vendor names, no "generation failed" notes, no costs you actually paid.
  • No ad ending. Don't pitch a service in the last third — videos that did performed badly. End on a copyable instruction card ("截图保存") plus a brand-free question for the comments.
  • Don't measure speed on a bad network. Quote official price and speed claims and attribute them ("官方说"); don't present your own timings as facts.
  • Judge honestly. Look at every generated image and say what is weak, not only what is strong. Crop anything resembling a real brand or store sign.

Flow

  1. Verify. Check every claim against the official site, docs, changelog and pricing, plus community threads. Keep RESEARCH.md with sources (example in the worked example). If a page returns 403, cross-check through coverage and say so.
  2. Format: 1080×1920, 60–90 s, unless the user says otherwise.
  3. Showcase assets: python3 -I scripts/gen_image.py assets/<name> --prompt "..." [--size 2:3] [--ref img.png] [--gif 12:150] [--markup layer.png]. Results and billing go to assets/results.jsonl (for you, never for the video).
  4. Script: script.json as {"scenes":[{"id":"s01","vo":"..."}]}. About 6 characters per second at speech_rate 8 (520 characters ≈ 85 s). End with the CTA question. Show the script to the user before TTS — TTS is billed.
  5. Voice: python3 -I scripts/tts_doubao.py script.json build/vo [ids] [--dry-run]. Writes <id>.mp3 and word timings <id>.json.
  6. Render: follow references/render-pipeline.md.
  7. Deliver: path, duration, size, measured LUFS and peak, which stills were checked, and the 1–2 weakest shots with reasons. Ask before any billable regeneration beyond what was agreed.

Individual skills in this repo

This repo contains 1 individual skill — each has its own dedicated page.

Related Skills