showcase-video-processor

Skill: showcase-video-processor

Overview

This skill is designed to help AI Model Teams efficiently produce high-quality showcase videos. Through a standardized pipeline, it transforms raw code, prompts, screenshots, and screen recordings into draft assets that match the "Luxury Tech" aesthetic, and ultimately organizes them into a complete showcase set.

Keywords: video processing, FFmpeg, model showcase, storyboard planning, asset organization

Core Features

1. Raw Asset Organization

Renaming: Follow the {YYYYMMDD}_{类型}_{描述}_{序号}.{后缀} convention.
Sidecar Documentation: Automatically create a same-named .txt description file for each asset, recording original parameters, recording context, and editing intent.

2. Draft Video Processing

FFmpeg-based lossless/high-quality processing, including:

Crop: Remove system UI (Dock, Menu Bar).
Trim: Precisely remove redundant material at the beginning and end.
Time Warp: Speed up the inference process while keeping result displays at normal speaking pace.
Zoom & Freeze: Keyframe close-ups and visual holds.
Quality Assurance: Preserve the original resolution and bitrate to the greatest extent possible, avoiding blur caused by re-encoding.
[Practical Case Reference]: See concepts/raw_assets_processing.md in FFmpeg 综合滤镜处理 (Crop + Partial Speedup)。

3. Multi-Granularity Storyboard Planning

Based on the design language under the concepts/ directory, write documents at different levels:

Micro Storyboard: The 4-step details for Single Case .
Meso Storyboard: The layout and pacing for Collage or Split Screen .
Macro Storyboard: The complete narrative structure for Showcase Set (core announcement -> capabilities -> metrics -> business).
[Practical Storyboard Example]: See references/example_releasing_model_max_storyboard.md for a complete, highly executable (precise to seconds and actions) multi-module release video storyboard reference.

The Pipeline

Phase 1: Ingestion

The user provides an asset directory.
The Agent scans and renames files, and generates description documents.
The user adds details in the description documents (e.g., "needs 4x speedup here").

Phase 2: Drafting

The Agent generates FFmpeg processing commands based on the description documents or user instructions.
Execute processing to produce clear, compact "draft video assets".
Verify the processing results (resolution, duration, retention of key content).

Phase 3: Compositing

Based on the draft assets, plan the storyboard documents.
Propose compositing suggestions according to the selected mode (single case, collage, split screen, comparison).
Finally integrate into a showcase set for release.

Best Practices

Quality First: Explicitly specify high-quality encoding parameters in all FFmpeg operations (such as crf 18 or prores）。
Sense of Rhythm: Accelerating the output (Output) is strictly prohibited; give the audience sufficient time to evaluate.
Luxury Tech: All annotations, borders, and transitions should follow minimalist, high-end visual standards.

Common Pitfalls

Over-cropping: When cropping, reserve a certain margin to prevent incomplete display on screens of different aspect ratios.
Encoding Loss: Avoid frequent format conversions.
Information Redundancy: The storyboard document should focus on "why it is displayed this way" rather than simply recording "what was done".