Quick walkthrough of the portal.

Generate lifestyle images

Upload a table lamp. The AI matches it to fitting interiors from that niche's library and creates the images with Nano Banana Pro.

Lamp image

Format

Quality

How many

Real size (optional)

Lamps per image:

Auto 1 2 3

  Different interiors each time Allow reused backgrounds reset used 📚 Pick interiors from library

Suggested

Generate

Bulk Lifestyle

Upload lamp photos, tick your saved prompts, and Nano Banana makes every photo with every prompt. Batches run on the server: start as many as you like, switch tabs or close the page, they keep going.

Variant images

Upload the interior you want (image 1), then the lamp you want in it (image 2). It keeps the interior exactly the same and swaps in your lamp — pick the lamp type below so the instruction is tailored (table lamp / pendant / ceiling light…). Add several lamps to make a set at once.

① The interior you want (your example scene)

② The lamp(s) to put in it (pick one or many)

Multi-head variants

Turn single-head lamps into a multi-head version. Upload ONE example of the multi-head light you want (image 1): a 3-head pendant bar, a round ceiling plate with 4 heads, a cluster, anything. Then add your single-head lamps (image 2). Each single lamp comes back as that exact fixture, in the same photo: same mount, same number of heads, same layout and drop lengths. Only the heads become your lamp.

① Example multi-head light (the build to copy)

② The single-head lamps (pick one or many)

White background with ChatGPT

Upload supplier photos (up to 40 at once). For each photo Claude writes a product-photo prompt with your own instructions (one prompt per colour when a photo shows a variant set), then your own ChatGPT makes a clean 1:1 photo on pure white with the supplier photo as reference. Four photos are made at the same time, about a minute each.

Image recreator

Upload photos in bulk. Each one comes back as a clean square (1:1) photo of 2048 × 2048: all text (also logos, watermarks and dimension lines) and all people removed, sharper and cleaner, and everything else exactly the same. Tall or wide photos are extended to a square, so nothing is cut off. Every result is checked automatically and made again when something is wrong (up to 3 tries per photo, the best one is kept). About a minute per photo, four at a time.

Multishot

Upload 2-8 different lamps of the selected type and Multishot stages them together in ONE scene — picked automatically from that type's library to fit the set. One collection shot per result, every fixture an exact copy of its photo. Supported: Table, Pendant, Ceiling, Wall, Outdoor wall, Outdoor floor. Switch type with the selector above.

UGC pictures

Upload a lamp and get photos that look like real customer photos (UGC). For pendants, chandeliers, ceiling and wall lights the room is a real photo from our interior library that is kept exactly as it is: we find the light that already hangs there, paint yours in its place and put the original photo back around it, so the room, the light, the height and the size are real. The other shots (unboxing, installing, holding, just installed) are made around that same home. People are mixed automatically. Every run picks other rooms and moments, and rooms already used for this product are not used again.

Interior library — Table lamp

Each lamp type has its own library. Filled by the scraper — this view is read-only. Switch type with the selector above.

Interior judge

Photos from architonic.com product pages, one at a time. → Yes adds the photo to the library of the chosen lamp type, ← No skips it. Nothing is shown twice.

Trial accounts

Temporary logins for trial candidates. Each one has its own password, only the tabs you give it, a maximum number of images per day and an end date. The server enforces every limit. Resetting the password or switching the account off logs the person out at once.

Accounts

Images are counted per UTC day (midnight UTC = 8 pm in Venezuela, 2 am in the Netherlands). A failed image does not count.

Names

People who log in with the team or trial password type their own name. A name in Staff is always spelled that way (also in other capitals or with one letter wrong), names like a model name or "test" are refused, and a merged spelling counts as that person everywhere in this tab, also in the past.

Every call

Newest first. Click a row for the prompt excerpt, the token split and the error text if it failed. Failed attempts are listed too (they cost nothing but Google counts them).

Before the ledger (history)

The ledger starts on 18 September 2026 and has no names before that. Older jobs still sitting in the job store can be counted: images vs vision calls per day, with an ESTIMATED cost (a 2K image at the Nano Banana Pro rate, a judge-style vision answer at the gemini-2.5-pro rate, other vision at the flash rate).

Settings

Log in once with the team password. The AI keys live on the portal's server and never touch this browser — nothing to lose, nothing to steal.

Your name is attached to every image and vision call you make, so the owners can see who creates what and what it costs. Admins log in here too, with their own password.

Every tab uses the selected AI with the same prompts. Heads up: ChatGPT makes the images on your own ChatGPT subscription on your Mac (Leon and Egbert): about a minute each, up to about 1.5K (no true 2K/4K), with their own look.

If image generation errors, copy the red message and send it over — it usually means one model id or setting needs adjusting.

Generation prompt — Table lamp

This is the exact instruction the AI uses to place a table lamp. Each lamp type has its own prompt — switch type with the selector at the top. Edit and Save, or Reset to the default for this type.