🔁 Text-to-image vs image-to-image, explained
Same models, different jobs. Text-to-image invents a scene, image-to-image transforms one. Confusing the two is the single most common reason creators waste credits.
Text-to-image — invent from scratch
You describe a scene in words, the model builds it from zero. Perfect for concepts, moodboards, blank-page ideation, and anything where the subject is imaginary.
Image-to-image (edit) — transform what you have
You upload an image, the model treats it as the anchor and modifies it based on your prompt. Perfect for restyling, background swaps, wardrobe changes, product-shot lighting and character consistency.
How to spot the right job
If keeping the exact subject matters (a real person, a real product, a real place), use an edit model — Nano Banana edit, Flux Kontext or Seedream edit. If the subject is imaginary or interchangeable, use text-to-image.
Prompt style is different
Text-to-image prompts describe the whole scene. Edit prompts describe ONLY the change ('swap the background to a minimalist grey studio, keep everything else identical'). Overwriting a prompt on an edit model is the #1 way people lose the subject.
Ready to put this into practice?
Open the studio and apply what you just learned in under a minute.
Try both in the studio