Key instructions to preserve
No beautification, no face reshaping, no expression change unless stated.
IDENTITY LOCK: Use the uploaded photo as the sole identity reference. Preserve the subject's facial structure, skin tone, eye shape, hairstyle, and every distinguishing feature exactly — the person must remain instantly recognizable. No beautification, no face reshaping, no expression change unless stated. SCENE: Subject depicted as a divine angelic figure descending through a break in the storm clouds over a modern city at dusk — massive white feathered wings fully extended behind them catching the final rays of golden light, while below them the city lights come on one by one. Their own clothing is unchanged — no robes, no costume. PHYSICS & CRAFT: The golden-hour light source is the setting sun breaking through the clouds at 1800K — pure gold rim light on the wing edges and face, the inner wing surfaces lit by diffuse sky at 6500K. Individual feathers on the wings are physically rendered — rachis, barbs, the slight translucency at the tips where light passes through. City below in blue-hour dusk, windows beginning to glow warm. TECHNICAL EXECUTION: 35mm f/2.8, looking up from street level, the subject and wings framed by converging city buildings, the break in the clouds creating a natural vignette, the city lights blurring into abstraction below. HARD CONSTRAINTS: The subject's face must be clearly recognizable. The wings must be physically plausible — real bird feather structure at this scale. No halo, no religious iconography. The modern clothing is essential to the concept.
0 people have copied this prompt
Angel Descending Over City at Dusk is a dark fantasy prompt for GPT Image 2, composed for a 4:5 image. No beautification, no face reshaping, no expression change unless stated.
No beautification, no face reshaping, no expression change unless stated.
Keep the prompt structure, then replace the main subject, action, or environment in "No beautification, no face reshaping, no expression change unless stated.".
Adjust one visual signal at a time, such as "35mm", "f/2.8", "rim light", "ai image prompt", so you can identify which instruction changes the result.
Keep the focal subject prominent in the portrait frame while preserving enough background detail for feed and poster layouts.
Use one clear reference image and retain the identity-lock instructions when changing clothing, lighting, background, or camera direction.
Use direct natural-language edits after the first generation. State what must remain unchanged before describing the single element you want to revise.
Judge the result by how clearly it communicates the intended dark fantasy direction while keeping the subject, environment, and visual treatment consistent.
Three simple steps to generate your AI image.
Click the "Copy Full Prompt" button above to copy the entire prompt text to your clipboard instantly.
Open your preferred image generator and paste the prompt into the text input. You're ready to create.
Hit generate and watch the magic happen. Tweak the prompt parameters to make it your own.