image-gen · PromptZap Team

Consistent Character Anchor

Solves AI art's biggest weakness — the same character across a series — with a reusable verbatim anchor block plus --cref reference technique.

Framework

standard

Model targets

claude, chatgpt, gemini

Prompt text

I need the same character to appear consistently across multiple images. Define a fixed "character anchor block" I'll reuse verbatim in every prompt:

[Age/apparent age], [hairstyle + exact color], [face shape/distinctive features], [exact clothing + colors], [any accessory that's part of their identity], [consistent art style, e.g. "flat vector illustration, muted earth-tone palette"].

Then for each new scene, keep the anchor block word-for-word identical and only change the action/setting/pose portion of the prompt: "[anchor block], now [new action], in [new setting], [lighting for this scene] --ar [ratio]"

If using a tool with image reference support, also attach the first generated image as a style/character reference (--cref or equivalent) alongside the text anchor for maximum consistency.