My Process for Creating Realistic AI Images

Step-by-step Breakdown

Prompt:

Vertical 9:16 photo with a vintage fisheye look and harsh flash. Inside a tiled Amsterdam metro station, two young people take a mirror selfie in a large convex security mirror above the escalator.

The girl up front holds her phone in front of her face. She's wearing rectangular black sunglasses, a denim jacket over a white tee with the OpenOs logo, loose light-wash jeans, and white sneakers. Her nails are long, glossy, and painted with red flame tips. She’s wearing chunky silver rings and layered necklaces.

Behind her stands a guy in a cream hoodie with “AMSTERDAM” on the front, under a faded blue oversized jacket. He’s also in loose jeans and sneakers. He holds a shopping bag and a lime green drink. His face is half hidden by the hood and a bit of motion blur.

Strong flash lighting hits the shiny subway tiles and the metal edges of the mirror. The vibe is bold, playful, and a little chaotic—like a Gen Z-style fashion snapshot from an underground evening out.

1. Aspect Ratio and Format

 “I use a 9:16 vertical photo.”
 This sets the frame the way I want it — vertical, portrait mode.
 Perfect for Instagram, Reels, or TikTok.
 Also helps with styling shots that feel like fashion or street content.

 
2. Lens & Lighting Effects

 “Shot with a vintage fisheye lens and strong flash.”
 This makes it feel real.
 The fisheye lens warps the edges and adds perspective — great for a mirror look.
 Flash lighting blows out the highlights, like an actual underground photo.
 It also makes skin pop more naturally.
 

3. Scene Context

 “Inside a tiled London Underground station.”
 Naming a real place helps.
 Models know what the London Underground looks like.
 Saying “tiled” gives a material the AI can latch onto.
 

4. Action and Framing

 “Two stylish young adults pose for a mirror selfie in a big convex mirror above the escalator.”
 This explains the setup.
 Convex mirror = distortion
 Escalator = setting
 Stylish = vibe, without over-explaining
 

5. Foreground Subject Description

 “The girl in front holds up her phone, half covering her face.”
 I always define who’s in front.
 Partial face coverage feels real.
 It also shapes how the photo is framed.
 

6. Outfit and Detail Layering

 “Black rectangular sunglasses. Denim jacket over a white Supreme tee. Baggy light jeans. White sneakers.”
 Layered clothes help build realism.
 Denim + Supreme feels styled.
 Baggy jeans + sneakers match Gen Z streetwear.
 I add detail too: glossy red flame nails, chunky silver rings.
 These make it feel like an actual photo, not just a render.

7. Secondary Subject Description

 “Behind her stands a guy in a cream hoodie.”
 I don’t repeat the same style as above.
 He’s slightly back, partly blurred, hood over face.
 Holding an H&M bag and a lime drink — feels casual, like a real moment.
 These choices make the AI treat him like a real person.

8. Lighting and Texture Detail

 “Flash casts sharp highlights on glossy subway tiles and chrome mirror edges.”
 This is about how light works.
 Real flash = harsh reflections
 Glossy materials = reflected light
 Details like this add realism fast.

9. Mood + Vibe

 “The photo feels raw, playful, and fashion-forward — like a Gen Z editorial shoot underground.”
 This line tells the AI how to style it.
 “Raw”, “playful”, “fashion-forward” are specific and guide the mood.
 Streetwear campaigns use this tone a lot.
 

Final Tip

 
Think like a director and a stylist.
 You’re not just asking for a “cool” image.
 You’re building a photo shoot — frame by frame.
 Each sentence should guide a choice: angle, light, pose, detail.
 
Pro tip:
I use ChatGPT to polish my prompts.
I also Google photo styles if I’m not sure what look I want.
That mix works well — and experience helps too.