Only the background changes. Your subject photo stays as it is while the AI background generator builds a new scene behind it. The canvas carries 34 image models, so a result you do not like is one retry away on another model.
A background you cannot re-run on another model, or re-open next month, is a background you will end up making twice. Most tools in this category put one model behind the button and stop at the download.
The steps below assume you already have a subject photo, because compositing is the harder of the two jobs.
Drop a JPEG, PNG or WebP onto the canvas. The cutout node separates the subject from whatever is behind it and hands you a clean edge to work with.
Write the scene, then set model, aspect ratio, batch size, resolution and file type in one panel. Type / for a saved prompt template, or @ to reuse an element.
Bring the subject and the background together on one node, check the contact shadow where they meet, fix the edge with the mask brush, then export.
The list runs from what goes into a generation to what happens after the download.
Describe a room, a street, or a flat colour field, and the generator returns a background image you can composite into later.
Generate a backgroundIdeogram V3, Gen4 Image, Omni Image and FLUX Kontext Multi read reference images, so you can point a generation at an existing campaign frame instead of describing it in words.
Generate a backgroundThe same scene comes out at 1:1, 4:5, 16:9, 9:16 or 21:9, so a listing image, a Story frame and a wide banner share one look.
Generate a backgroundThe canvas keeps the model choice, the prompt and the option panel. The next SKU or the next episode starts from those instead of a blank box.
Generate a backgroundBatch generation returns as many as nine images from a single pass, which is usually enough to find the one that matches the subject lighting.
Generate a backgroundBria RMBG 2.0 for the cutout, a mask brush for the edge, crop for the placement, Topaz or Magnific when the export needs more pixels.
Generate a backgroundCopy a prompt and swap the nouns to fit your own scene. Each one ends with the framing it was written for.
A homepage hero and an email banner are far wider than anything a camera framed, and cropping a normal photograph to 21:9 throws most of the scene away. Generating the background at that shape from the start leaves the subject sitting where the layout needs it.
Generate a background
A remote team sends in phone selfies taken against five different walls. Each keeps its face and gets the same backdrop, so the about page reads as one set of photographs. For portraits generated from scratch, see `/features/ai-headshot-generator`.
Generate a background
One campaign has to land in a feed post, a Story and a display banner. Cropping a single photograph into three shapes tends to push the product out of the safe area, so each placement gets its own backdrop.
Generate a background
A director blocking a sequence needs the location before the cast is booked. The background stands in as a reference frame, so the shot list can be argued over before anyone spends a day on set.
Generate a background
Most marketplaces want the main listing image on a plain white field, and a shot taken on a kitchen counter will not pass that check. The subject is already in the frame, so only the backdrop is regenerated as a white sweep at the ratio the listing asks for.
Generate a background
A seller whose whole catalogue was shot on a kitchen table cannot book a studio day per SKU. The subject photos already exist, so you only regenerate the backdrop, and the same white sweep runs down the list.
Generate a background
| Input | A subject photo, or nothing at all when you are building a background from text | |
|---|---|---|
| Export format and resolution | PNG or JPEG, at 1K, 2K or 4K | |
| Aspect ratios | 1:1, 4:5, 5:4, 3:2, 2:3, 16:9, 9:16 and 21:9, plus wider and taller options such as 4:1 and 1:8 | |
| Free plan | 40 credits a month, no card needed; results are for personal exploration only and are not licensed for commercial use | |
| Models used for compositing | FLUX Kontext and Qwen Image Edit for single-image edits; Nano Banana 2, Qwen Image Edit Plus and Seedream v4.5 Edit also accept multiple images | |
| Batch size | Up to 9 images per run. A 4-image batch costs roughly four times a single image at the same settings | |
| Credit cost | Varies by model and resolution. The exact cost is shown on each node before you run it | |
| Commercial use | Paid plans only, with per-model exceptions. See /help/commercial-rights/can-i-use-generated-content-commercially | |
| Uploaded images | Retained while the project exists. Used to operate Astorie and to train and evaluate its own and affiliated models, unless an enterprise agreement says otherwise. Workspace projects are visible to the workspace |
All 34 image models are available from the same node, so the choice costs a click rather than a subscription. These five come up most often on background work.
It is a tool that creates the setting behind a subject instead of deleting it. Most of the category does one of two jobs: replacing the background in a photograph you already have, or producing a background image on its own. The first has to match the lighting already on your subject; the second does not.
Yes, and it does not ask for a card. A new account starts with 40 credits a month, which is enough to generate backgrounds and compare models on the canvas. Results made on the free plan are for personal exploration only and are not licensed for commercial use, so treat it as a way to see the output before you pay. Prices are on /pricing, and each node shows its credit cost before it runs.
On a paid plan, yes, and ownership of the output transfers to you the moment it is created. The free plan is for personal, non-commercial evaluation. Two limits hold on every plan: a model labelled non-commercial in its node panel stays non-commercial, because Astorie cannot override a provider's own terms, and clearance for real-person likenesses, brands and copyrighted characters is still yours to get. See /help/commercial-rights/can-i-use-generated-content-commercially.
Decide the direction first and hold it in both halves. Write it into the background prompt, camera left or overhead or backlit, and check it agrees with the light already on your subject photo. Merge the two with an editing model such as FLUX Kontext rather than pasting one over the other, and ask for a contact shadow.
Open the canvas, pick a model, and run the first background against your own subject photo.
