Lanxi Script Character Asset Extraction

Problem

In dynamic comic production, character consistency is often the hardest part. The same person may appear under a formal name, alias, kinship term, or group label, while also requiring variants for costume, age, disguise, or combat state. Manually building an asset table can miss variants, and raw image prompts can drift toward storyboards, scene art, or copyrighted styles.

How It Works

This skill targets design-media workflows and converts script text into character assets for AI dynamic comics. It expects a script, scene breakdown, plot summary, dialogue, or narration, plus a project-wide style prompt. If the style prompt is missing, it asks for one or marks the result as a temporary neutral draft.

The main steps are: reading character fields, speaker tags, action lines, dialogue titles, and group references; building a character mention table; merging aliases into one asset name while preserving original labels; ranking assets as S, A, or B level, not-yet-appearing, or variants; defining visual anchors such as face shape, features, hair, color system, accessories, scars, and material language; and adding controlled art fill-ins when the script lacks appearance detail.

The output is not scene art or storyboard prompts. It produces standalone, copy-ready Chinese GPT-image 2 character asset prompts. Variants keep identity stable, background roles remain reusable, and prompts avoid named artists, films, games, celebrities, or copyrighted style references.

Boundaries

This skill is for extracting character assets and character prompts from scripts, not for full storyboards, scene design, camera movement, or complete short-video production. Without a project style prompt, results should be treated as drafts. Minor characters are handled age-appropriately and non-sexualized.