Opus 5.5 is good at explainer videos(launchvideo.io)
407 points by iacguy 19 days ago | 214 comments
tl;dr: launchvideo.io uses Claude Opus 5.5 to generate explainer videos from a URL or prompt in about 4 minutes, consuming ~100k tokens per video. The system runs as a single serverless agent on OpenComputer with three tools (web_fetch, check_scene, render_video): the model writes the film as HTML/CSS/JS, then a headless Chromium in a fresh microVM deterministically renders frames to 1080p30 via ffmpeg—no video model involved. The entire pipeline is one TypeScript agent file, deployable in one click.
HN Discussion:
  • ↑Sharing impressive external examples of Opus 5.5 output that reinforce the article's claims
  • ↓The generated explainer videos are low-quality, low-effort content with little real value
  • ↓The demo videos shown aren't as informative or impressive as claimed and repeat points
  • ↓Skepticism about model release demos being cherry-picked or misrepresented
  • •Practitioners building similar tools share related experience and interest in the approach