GPT Image 2.0 generates realistic night car selfies with phone-screen framing

A new technique for generating hyper-realistic AI selfies with GPT Image 2.0 has emerged on Reddit, focusing on indirect framing to mimic candid iPhone photography. The prompt places the camera from the passenger seat, showing only a woman’s hand holding a black iPhone Pro horizontally. The subject’s face appears solely on the phone’s front camera preview, complete with UI elements like focus frames and camera buttons. This framing choice shifts the image from a typical AI portrait to a believable “captured by accident” lifestyle shot.

The prompt specifies cinematic shadows, high-contrast low-light conditions, warm blurry city lights outside the window, and realistic imperfections like slight motion blur and noise. The woman in the preview touches her lower lip with a black lip pencil, adding a subtle editorial detail. The result is a 9:16 ratio image that feels like an authentic night car selfie. This approach demonstrates how carefully structured prompts can overcome AI’s tendency toward overly polished or unnatural results, opening new possibilities for realistic AI-generated lifestyle photography.