Open-source B-roll skill turns scripts into 5-second vertical clips with Codex and Gemini
yangyi · x · 2026-07-21
This open-source B-roll skill turns a voice script into a stop-motion-style vertical clip through a gated pipeline:
- install and self-check the Codex skill
- compress a script into a visual metaphor
- generate and confirm a colored collage still plus contact sheet
- animate only after the still passes review, using Gemini Omni Flash for start/end-frame assembly
- remove audio and export a 9:16, 5-second clip for voiceover backplates
- run three QA checks: frame-by-frame sampling, empty-scene verification, and end-frame comparison
The author says the workflow is intentionally designed to avoid expensive reruns, and notes it currently works only on Codex because it depends on built-in imagegen, with Claude Code and others still being adapted.
More from coding & agent
- A 200-line agent design treats the whole conversation as a single file — Low-Sandwich1194 · 2026-07-21
- SakanaAI adds W&B logging to ShinkaEvolve for evolutionary search tracing — _ScottCondron · 2026-07-21
- An intern says 1M tokens still do not make agent handoffs obsolete — ke1lle · 2026-07-21
- An agent-loop guide says good systems need eight separate exit conditions — blaizedsouza · 2026-07-21
- Supabase ships official plugins for Kimi Code and Kimi Web — dshukertjr · 2026-07-21
- Belgie: Embeds a TypeScript Sandbox for Python AI Agents Without Node.js — TheRealMrMatt · 2026-07-21