Core technique
The prompt uses a reference-controlled workflow: identity information comes exclusively from the personal image, while visual characteristics are transferred from the reference scene. It combines a three-quarter seated composition, 85mm-style perspective, f/1.8–f/2.2 depth of field, window illumination, warm café ambience, and controlled background bokeh. Facial realism is reinforced through explicit anatomical, skin-texture, hair, and glasses requirements.
Recommended base models
Flux-based models should handle the detailed photorealistic environment, clothing, and natural materials well, although identity preservation depends heavily on the specific identity-reference workflow. GPT-Image-1 is well suited to following the long identity, environment, pose, and negative constraints together. SDXL-based workflows can produce the café lighting and 4:5 portrait composition effectively, but may need stronger face-reference or identity controls to maintain recognizable facial structure. Midjourney can produce the editorial atmosphere and café styling strongly, but the exact facial identity requirements may be harder to maintain consistently than in identity-focused workflows.
Suggested negative prompt
different person, identity loss, face replacement, generic model face, altered facial proportions, changed eyes, changed nose, changed lips, changed jawline, facial beautification, plastic skin, waxy skin, excessive smoothing, distorted glasses, warped lenses, duplicate glasses, excessive lens glare, artificial eyes, malformed ears, distorted neck, unnatural shoulders, stiff seated pose, deformed hands, extra fingers, unrealistic clothing folds, artificial hair, excessive hair shine, overexposed window, blown highlights, crushed shadows, excessive HDR, oversharpening, excessive bokeh, cluttered café background, neon colors, oversaturated skin, teal-orange grading, logo, text, watermark, CGI, 3D render, AI artifacts
Variation notes
Reducing the identity instructions would likely make the prompt easier to follow but increases the risk of facial drift. Increasing the aperture effect beyond the specified f/1.8–f/2.2 look could make the café environment too blurred, while keeping the background moderately visible better supports the requested lifestyle-editorial atmosphere.
