Before-and-After Image Prompts: Headshots, Rooms, and Product Shots

Before-and-after images fail when the "after" is a new picture that happens to include the same person, sofa, or bottle. Clients can feel the swap even if they cannot prove it. The job is a controlled edit: keep identity, change one job-shaped variable.

That is a prompting problem as much as a model problem. Whether you are using an edit-capable image model, a Midjourney reference, or a Flux img2img pass, the prompt has to say what is locked and what is allowed to move. If you only describe the desired after state, the model will invent a new after.

This walkthrough covers three jobs that show up constantly in freelance and in-house work: headshots, rooms, and product shots. Steal the structure, then save the variants you like in PromptDig (Browse more prompts) so you are not rebuilding lock lists from scratch.

The lock list comes before the glow-up

Write two blocks every time. The first block is sacred. The second block is the job.

Sacred (do not change):

• Identity: face, body type, product silhouette, room architecture
• Camera: angle, lens, crop, distance
• Lighting direction, unless lighting is the job
• Brand colors and logos, unless the job is a rebrand

Job (change only this):

• Wardrobe, backdrop, or retouch level for a headshot
• Paint, furniture, and clutter for a room
• Surface, scale, or context for a product

A useful prompt says both in that order. Models follow the last exciting adjective if you are not explicit. Put the locks first.

A skeleton you can reuse:

This is an edit of the uploaded image, not a new image.
Keep: [identity / architecture / silhouette]. Keep the same camera
angle, crop, and lens. Keep lighting direction.
Change only: [one job].
Do not add people, extra products, or new windows.
After state: [concrete description].

If your tool has a strength, denoise, or edit-intensity slider, start low. You want the after to still be the before with a job applied. High strength is how a "new paint color" becomes a new apartment.

Headshots: lighting and wardrobe, not a new person

The usual request is "make this LinkedIn photo look professional" or "studio, same person." The failure mode is a neighbor who borrowed the subject's cheekbones.

Lock the face explicitly. Name what must stay: face shape, eyes, hairline, skin texture, glasses, facial hair. If you want a retouch, define the retouch as a ceiling: even skin, not porcelain; stray hair reduced, not a new haircut.

Then name the job in production language:

Edit of this headshot. Same person, same face, same hairline, same glasses.
Same camera height and crop (head and shoulders).
Change only: replace the busy office background with a seamless
medium-gray paper backdrop. Soft key from camera left, gentle fill,
no beauty-dish glare. Neutral professional wardrobe, navy crewneck,
no logos. Natural skin texture. Not a different person, not younger.

For a wardrobe-only after, say "same background, same light, change clothing only." For a lighting-only after, say "same wardrobe and background, move to window light from the right." One variable per generation. If you need three afters (studio, outdoor, dark editorial), run three prompts. Combining them is how identity drifts.

If you have a reference of the lighting you want, use it as style or scene reference and keep the original photo as the identity reference. Do not mash both into one vague "like this" sentence.

Rooms: architecture is the product

Real-estate and interior afters go wrong when walls move. The model "improves" the space by adding windows, raising ceilings, and inventing a second sofa.

Lock the bones: window count and placement, door, ceiling height, camera standing point, lens. Then stage.

Edit of this living room photo. Keep the same walls, windows, ceiling,
floor, and camera angle. Do not add windows or change room size.
Change only: repaint walls warm off-white, replace the worn gray sofa
with a mid-century walnut-legged sofa in camel fabric, add a simple
oak coffee table, reduce clutter on surfaces. Same daylight from the
existing windows. Lived-in, not a catalog showroom. No extra rooms
visible, no people.

"Lived-in, not a catalog showroom" is doing real work. Empty perfection reads as fake. Keep one honest object (a book, a throw) so the after still looks occupied.

For renovation sequences, generate a chain: empty + paint, then furniture, then styling. Each step uses the previous after as the new before. Jumping from a messy room to a fully designed interior in one hop is how floor plans change.

Product shots: one bottle, two worlds

Packshots and lifestyle shots are different cameras. Do not ask for both in one prompt.

Packshot after (ecommerce):

Edit of this product photo. Keep the exact bottle silhouette, label,
and cap. Keep brand colors and typography on the label readable.
Change only: place the bottle on seamless white, centered, soft
shadow under the base, even studio light, no props, no condensation
unless it is on the original. Straight-on, mid-length, catalog crop.

Lifestyle after (ads and PDPs):

Edit of this product photo. Keep the exact bottle and label.
Change only: set it on a sunlit kitchen stone, one citrus slice
outside the bottle's silhouette, shallow depth of field, still
the same bottle size in frame. Do not reprint the label. Do not
add extra bottles.

The phrase "do not reprint the label" matters. Models love to hallucinate cleaner type. If the label is the legal design, treat it as sacred and composite type in a layout tool when you need a new SKU.

Across all three jobs, review the after as a diff, not as a pretty picture. Zoom on faces, window counts, logos, and seams. If two things changed, you asked for two jobs. Run it again with a tighter lock list.

When a prompt finally holds identity, share that lock-list version (Share a prompt). The after adjectives are easy to rewrite. The locks are the asset.