I’ve been trying image generation in Grok and want to refine a result over a few prompts rather than start over each time. If I ask for a different background or a small change to one object, does it usually preserve the original framing and the other details?

Curious how well follow-up image edits work in practice, especially when the image includes text or several objects that should stay put. Do you get consistent revisions, or is it more dependable to regenerate and compare?