Mobile prompt for the 3 frame urban headshot (same wavy bob, gray cropped blazer with front cutout, white wide leg pants w/ gold chain belt). Include language to help keep skin texture and fabric detail natural. Background blur should be uniform. Points out where outfit / face consistency often fails.
photorealistic lifestyle portrait collage of the same woman with shoulder-length wavy brown hair, natural makeup, wearing a charcoal gray cropped blazer with 3/4 sleeves and small twist cutout at the front hem, high-waisted white wide-leg trousers with thin gold chain belt, delicate gold necklace and hoop earrings, soft natural daylight, realistic skin texture with visible pores, shallow depth of field with blurred city background, three panels: top half-body smiling on city street, bottom left full-body standing with hands in pockets, bottom right seated on rooftop bar stool holding a martini glass, consistent face and outfit across all panels, fabric weave detail, natural shadows -- AI image prompt
For the gray cropped blazer with twist, small horizontal cut above waist. With high waisted white wide leg pants and thin gold chain belt. You need to have it exactly same in three different poses. That's the tricky part here not just one photorealistic face.
Soft overcast daylight on the street (on frames) and warm late afternoon sun on roof. Natural skin texture, no waxy smoothing of skin. Blazer looks like nice matte wool with close weave. Pants look like smooth crepe that hangs but doesn't stick to body. Chain belt is one strong highlight point. Martini glass and tall mint highball in rooftop frame show clear transparency and thin rim reflection. Not solid plastic looking glass on roof. Oof background (traffic lights bokeh circles up top), blurriness of glass towers on left, warm sky/metal railing right side.
What the prompt needs to lock in
Describe the person once, then repeat that exact outfit description in each panel. Use specific nouns: "shoulder length wavy brown hair" "cropped charcoal gray blazer with 3/4 sleeves and front twist cutout" "high waisted white wide leg trousers" "thin gold chain belt" "gold hoop earrings" "delicate gold necklace" "strappy metallic heels". And make sure to add "same face, same outfit in all panels". Then for the framing, describe that it's a collage: "three panel collage. top half body shot. bottom left full body standing. bottom right sitting at rooftop bar table." This way it helps the generator hold on to identity rather than creating 3 different women.
For the photographic style, just be simple and studio clean (even though you're outside): "photorealistic. natural skin texture with pores visible. soft natural daylight. shallow depth of field. 85mm portrait lens. blurred urban background." The idea is to emphasize sharp fabrics/skin up front and blurry buildings in back... but not that the model physically traces light or anything like that. Just what you want to see visually.
Why this wording helps, and what to change when it drifts
By specifying the full outfit in each panel description (so it doesn't drift) you reduce opportunity for them to replace your blazer or pant cut. Specify material ("matte wool blend blazer, smooth crepe trousers") to ground sheen. If your blazer gets shiny and/or pants look like denim, use those material words instead of "elegant" etc
If the face changes from panel to panel, shorten background description and put "same woman. same facial features" first in prompt If gold chain belt isn't there for full body shot add: thin gold chain belt at waist (draped on left side) remove other competing accessories from that sentence
For background blur, "shallow depth of field, background blurred, bokeh lights" is usually enough. If the street stays too sharp, add "subject in sharp focus, background out of focus" and reduce city detail words. The approach used for street portraits with natural blur shows how little extra city description you need once focus is stated.
Limits to expect
Keep in mind that this is a simulation not an actual photo so there will be some minor inconsistencies. Hands (especially when holding the martini glass) can distort. The chain belt may vary in length and link size from one frame to another. The folds of the fabric won't line up exactly across panels. When you get the full body shot, facial features blur a bit more as does the weave on the blazer - it's normal for this distance.
We don't know the exact model and settings to get this image, so consider it a portable base prompt. Not an absolute reconstruction of what was used. The above changes are just some untested ideas that you can iterate on. They are not guarantees for getting the same results in different generators. If you want true identity consistency when working with clients - generate one approved half-body first and use this image as character reference for other two poses instead of text only.
Comments
Post a Comment