Keeping One Character Consistent Across a Whole Article's AI Illustrations

Описание и применение

Инструмент InkDoo генерирует последовательный набор иллюстраций для статьи, используя строгий шаблон подсказки, где персонаж описывается как точный, счётный спецификатор, выполняет ключевое действие и сохраняет единый стиль. Это устраняет разрозненные изображения и обеспечивает целостность визуального повествования.

Задача

Illustrating a long blog post with an image model often yields inconsistent art styles and multiple versions of the same character, making the post look disjointed.

Результат

InkDoo produces a cohesive set of illustrations with a single, well‑defined character, consistent style, and correct placement within the article.

Рабочий процесс

• 1. Planner LLM reads the article and outputs JSON with ideas worth illustrating, paragraph anchors, and scene descriptions.
• 2. Each scene is rendered by an image model using a strict prompt template that includes a literal character spec, a fixed style block, and minimal per‑image variables.
• 3. The exporter inserts the hosted image URLs back into the Markdown at the correct paragraph anchors.
• 4. Optional steps: switch characters via whole‑word replace, redraw single images, or adjust labels for language.

Доказательства

The biggest single win was writing each character as a short, very literal visual spec, then pasting that exact string into every image prompt.

Источник

The planner writes scenes in English with the character's name in them, so switching is a careful, Unicode‑aware whole‑word replace of the old name with the new one.

Источник

The exporter splits the article into blocks, normalizes whitespace, and matches each anchor to a block.

Источник