nickqi688/video-generation-workflow

Default language: Chinese Default language: Chinese

Turn a video requirement, a practical article, or a product introduction into an executable video production process: first clarify the goal and style, then break down the script and storyboard, generate a visual preview prompt, and finally use code-driven methods like HyperFrames / Remotion / HTML / React to generate renderable video scenes.

The core of this skill isn’t “writing a video prompt,” but helping Codex work like a video production project manager: asking questions while setting direction, breaking down and executing as much as possible, trying to break down aesthetic judgment, content judgment, and project implementation.

What is it suitable for?

• Turn the practical articles into tutorial videos
• Turn product introductions into product demonstration videos
• Change the tool tutorials to Xiaohongshu / Douyin / YouTube videos
• Generate storyboards, audio scripts, subtitle scripts, and visual preview prompts
• Guided Codex on replicating visuals and exporting videos using HyperFrames / Remotion / HTML / React
• Generates multiple background style directions to form a reusable background image library
• Provide the client with a video plan first, then fine-tune each shot based on customer feedback

Default output direction

When the user does not specify the frame, skill will first let the client choose:

• 16:9 横版: YouTube, Bilibili, product demos, courses, customer reports, default selection
• 9:16 竖版: Xiaohongshu, Douyin, TikTok, Reels, Shorts
• 1:1 方版: Information feed, ad preview, cross-platform reuse

If the user does not select it, the default is to press 16:9 横版视频.

Installation

Copy this repository to the Codex skills directory:

Afterwards, you can call Codex like this:

Or:

Workflow

1. Determine input type: short requirements, practical articles, product briefs, existing scripts.
2. Start by asking key questions: platform size, target audience, style references, narration style, available materials.
3. Generate a video brief: title, goal, duration, style, sound, risk points.
4. Split Shot: Each frame’s duration, visual tasks, subtitles, composition, animations, and sound.
5. Generate audio script and subtitle drafts: By default, output complete audio play, frame-by-camera audio play, and subtitle versions.
6. Generate visual rehearsal prompt: Set the visuals first, don’t rush to write code.
7. Preview Diagram Confirmation: Only after client confirmation can code animation proceed, unless automatic advancement is explicitly requested.
8. Code replication scenarios: text, numbers, UI, charts, and dynamic code implementation; Images are only used for reference, background, people, and product materials.
9. Single-lens preview and fine-tuning: Only the most critical issues are fixed each time.
10. Final assembly: Check duration, resolution, subtitle readability, audio-video synchronization, and export format.

Why release the visual preview image first?

Simply saying “sophisticated, tech-savvy, concise” is usually not enough, because AI interprets these words in many different directions.

This skill tends to have Codex generate visual rehearsal prompts for each key scene first, and if necessary, use the image model to create a reference image. After confirming the visual direction, the image is handed over to Codex to replicate as code animation.