Prompt: Turn the input character into a 90s disposable-camera flash snapshot: a candid, intimate photo, as if a friend pointed a cheap single-use camera at them and pressed the shutter.

Live-action translation: real skin with pores, real hair texture, real fabrics, real human proportions. Keep the character recognizable: hairstyle (buns stay buns, twintails stay twintails), hair color (unusual colors become dyed hair), eye color (natural irises or contacts), face shape, age and signature accessories. Everyday outfits become their real-world equivalent. Fantasy, armor or ceremonial outfits become 90s streetwear in the character's colors and motifs: a cape becomes an oversized jacket, armor trim becomes metal jewelry or belt hardware. Keep only the foreground character; anyone else becomes a blurred patron.

Framing: vertical 2:3, head to mid-torso, slightly off-center and imperfectly handheld. A pose that fits the place, such as turning back on a bar stool mid-laugh, leaning on a table over a drink, or singing into a karaoke mic.

The place: a small, busy izakaya at night by default, with handwritten menu strips, beer posters, paper lanterns and dim older patrons behind. The user may name another (karaoke booth, ramen counter, convenience store, train, bedroom). The flashed subject against the dim background carries the photo.

The camera: direct on-camera flash with hot highlights on forehead, cheekbones and nose, harsh falloff and a deep shadow behind the subject. Visible 35mm grain, the soft focus of a cheap plastic lens, a warm lab-print cast, heavy corner vignetting, and maybe a faint light leak or scan dust at one edge.

Mood: nostalgic and imperfect, like a photo someone found years later in a drawer.

Avoid: clean smartphone, HDR or DSLR looks; airbrushed or idealized skin; anime or cartoon rendering; garbled text beyond plausible shirt graphics and signage; watermarks or frames.

Real Model: GPT Image 2.5 Sunburst v2.5