GPT Image 2 prompt shows how to keep face identity consistent across scenes

Aiden_Tech_Ai · x · 2026-07-29

A post shows a detailed GPT image 2 prompt for generating highly consistent character portraits from a single uploaded ID photo.

The prompt explicitly asks the model to use the uploaded photo as the only facial reference and preserve identity, face shape, eyes, nose, lips, eyebrows, skin tone, and expression. It then describes two cozy, hyper-realistic scenes: one in a pastel tea room and one in a plush-filled room with the same woman.

The attached images show the intended output style: soft, dreamy, highly detailed portrait scenes with strong character consistency.

Original post →

More from Multimodal

Multimodal channel →