I’m trying Grok’s image generation for a set of illustrations that should look like they belong together. Reusing the same prompt gives me interesting results, but the style and details can shift from one attempt to the next.

Does anyone have a reliable way to keep a character or visual style consistent across several generations? I’m wondering whether prompt wording, reference images, or starting from an earlier generated image makes the biggest difference.

If there are limits to how much consistency to expect, I’d rather plan around them than keep rerolling blindly.