Same guy, same pose, same subway car, same people in the background. Only the outfit changed. This is the prompt: Replace only the clothing of the person with a realistic classic red-and-blue Spider-Man suit, unmasked, while keeping his face, hair, pose, hands, body proportions, background, car, lighting, framing, and every other detail exactly unchanged. Full photorealistic suit replacement only, with realistic fabric texture, black web pattern, chest spider emblem, and natural folds. Do not alter anything except the outfit. Why it works so well: “Replace only” tells the AI this is an edit, not a new image. Without it, most models regenerate the whole scene It lists everything that must stay: face, hair, pose, hands, background, lighting. The more you lock, the less the AI drifts “Unmasked” keeps his face visible. That’s the whole joke, and it’s what makes it look real Fabric details like texture, web pattern and natural folds stop the suit looking painted on It ends by repeating the rule: “Do not alter anything except the outfit.” Repeating the constraint really does help. Swap the outfit for anything: Batman suit, medieval knight armour, astronaut suit, Victorian gentleman, footballer’s kit. Keep the rest of the prompt exactly the same. Your turn: try it on a photo of yourself and post the result below. What outfit are you going with? 👇

All 6 of these photos are AI-generated, and each one has ONE thing that's physically wrong Some are obvious. Some took me a second look, and I made them Comment your answers like this: 1 = …, 2 = … and tell us your score out of 6 No peeking at other comments first Answers drop tomorrow

Just one boring phone photo and a better description of the scene The upgrade comes down to four things: 1. Surface: office desk → dark walnut wood 2. Light: flat overhead → warm side light from one direction 3. Story: add one detail that makes you feel something (the steam) Grounding: a soft shadow and reflection so the mug actually sits on the table That last one is where most AI product shots fail. If there's no shadow where the object touches the surface, it looks pasted in Try it: grab the most boring object on your desk right now and turn it into an ad. Post your before and after below. Stapler ads especially welcome.

Same prompt. Same woman, same café, same pose. I only changed one word: the light Overcast, golden hour, neon, candlelight. Lighting is the most underrated part of any AI prompt. It controls the mood, the colours and even how real the skin looks. Most people describe the subject in huge detail and then leave the light to chance Here's the base prompt, so you can try it yourself: Woman in her thirties with dark curly hair and a rust-coloured jumper, sitting at a small round table outside a corner café, holding a coffee cup, looking slightly off camera. 50mm lens, photorealistic. Lighting: [YOUR WORD HERE] Swap the last line and watch the whole story change Which one's your favourite? And what light would you try next: rainy window, harsh noon sun, car headlights?

Breakfast, sleepy dog, rainy high street, gym mirror selfie. Just someone's camera roll. Drop your guess below (1, 2, 3 or 4) before reading the comments.