Darin Broll

English | 简体中文

Turn a short spoken line, opinion, or abstract idea into production-ready prompts for an editorial paper-collage B-roll clip.

Darin Broll keeps the visual reasoning of the Ray Broll workflow while removing local media-generation dependencies. It guides the user through two approval gates, then delivers English prompts that can be pasted into external image and video tools.

What It Produces

The Skill follows one fixed sequence:

  1. Visual direction - meaning, emotion, visual metaphor, up to four object groups, color palette, and assembly order.
  2. First-frame and last-frame prompts - an empty paper-color opening frame and a completed editorial collage still.
  3. Video prompt - a locked-off, assemble-from-empty stop-motion sequence that ends on the approved composition.

The Skill pauses after steps 1 and 2. It does not spend generation credits or call media tools by default.

Why Use It

  • Converts abstract narration into one readable visual idea.
  • Uses a consistent editorial collage language: halftone photo cut-outs, flat paper color fields, cream keylines, and tactile shadows.
  • Limits each scene to two to four separable groups to reduce composition drift in video models.
  • Prevents fake text, logos, platform UI, watermarks, and unnecessary decorative objects.
  • Produces self-contained English prompts with no unresolved placeholders.
  • Supports single items, batches, and partial approval by item number.

Example Workflow

Start with a line such as:

I have run a digital-technology channel with more than sixty thousand followers.
I have published hundreds of posts, and I have made nearly every mistake already.

The Skill first returns only a visual direction, for example:

Core meaning: Experience comes from repeated publishing and repeated failure.
Visual proposition: A creator stands in front of an abstract channel page while video windows accumulate around them; the creator falls, then rises again.
Key groups: creator, abstract channel page, device cluster, video-window group.
Color: mustard yellow field with cream and burnt-orange accents.
Assembly order: channel page, creator and devices, video windows, fall-and-rise action.

After approval, it provides the two still-frame prompts. After those prompts are approved, it provides the video-generation prompt.

Installation

Clone the complete repository so SKILL.md, agents/openai.yaml, and LICENSE remain together.

Codex

git clone https://github.com/jidarin9383/darin-broll-skill.git ~/.codex/skills/darin-broll-skill

Claude Code - Global

git clone https://github.com/jidarin9383/darin-broll-skill.git ~/.claude/skills/darin-broll-skill

Claude Code - Project Only

git clone https://github.com/jidarin9383/darin-broll-skill.git .claude/skills/darin-broll-skill

For another Agent Skills-compatible client, clone the repository into that client's skills directory. Start a new session or restart the client after installation if the Skill is not discovered immediately.

Usage

Invoke the Skill explicitly:

Use $darin-broll-skill to turn this spoken line into an editorial paper-collage B-roll:
Trust me. I have been through this situation too many times.

Or in Chinese:

使用 $darin-broll-skill 给这句口播设计纸拼贴 B-roll：
相信我，这种情况我经历了太多。

Common confirmations:

确认
同意
OK
继续

Each confirmation advances only the current gate. Revisions remain at the current gate until approved.

Visual System

  • Vertical 9:16, approximately five seconds, silent by default.
  • One continuous locked-off shot.
  • Two to four large, separable paper groups.
  • Subject kept within the middle 70 percent with generous negative space.
  • Black-and-white halftone photographic cut-outs with restrained colored cardstock accents.
  • Flat uncoated-paper background, cream keylines, crisp cut edges, and soft paper shadows.
  • Piece-by-piece slide in, drop in, snap into place, fall, or rise motion.
  • No readable text, numerals, logos, watermarks, real platform UI, camera movement, morphing, or added objects.

See SKILL.md for the complete workflow, prompt contracts, color semantics, anti-drift rules, and failure recovery guidance.

Prompt-Only by Design

This repository does not include image or video API clients, model credentials, generation scripts, or local file requirements. The Skill writes prompts only unless the user explicitly hands the work to another generation tool.

That makes it portable across different image-to-video products and avoids coupling the visual workflow to one provider or model version.

Repository Structure

darin-broll-skill/
├── SKILL.md
├── agents/
│   └── openai.yaml
├── README.md
├── README.zh-CN.md
├── LICENSE
└── .gitignore

Updating

git -C ~/.codex/skills/darin-broll-skill pull --ff-only

Use the equivalent installed path for other clients.

License and Attribution

Released under the MIT License.

This Skill is adapted from gbro-collage-broll and its Ray Broll workflow. It preserves the upstream editorial paper-collage visual system while changing the execution model to external prompts and two human approval gates.

Original copyright: Copyright (c) 2026 狗哥笔记.

Modifications: Copyright (c) 2026 Darin Ji.

Keep the included LICENSE file and its copyright notices when copying, modifying, or redistributing this project.