Can I go without reference images?

Yes. A one-line idea is enough to plan and generate; the product set, on-model apparel, portrait and food spot templates need a reference image first.

Is the apparel model the same person in every shot?

Yes. Without a model photo the full-length front shot is made first and every other shot uses it as a reference, asking for the same face, hair and build; with a model photo the whole set uses that person. If the first shot fails the set stops and the other shots aren't charged; adjust and submit again.

Does AI planning cost credits?

No. Each person can plan up to 30 times per 24 hours; after that, fill in the plan yourself and generate directly.

Will the text on the images be wrong?

Only selling-point images and posters carry text, and it's the text you wrote, which the prompt asks to be reproduced exactly. AI can still draw characters wrong; redo that image if it does.

Am I charged for an image that didn't come out?

No. Each image is charged only when it succeeds; credits held for one that failed or was rejected are released.

How is the true white background made?

It isn't generated: a local segmentation model cuts the product out of the photo you uploaded and places it on pure white, so colours and details stay pixel-exact, and it's free. Photos with a clear subject on a simple background work best; if it can't be cut cleanly you're asked for another photo and the rest of the set is unaffected.

One-click AI images: product listings, on-model apparel, posters and portraits as a full set

What you get

One ZIP: images named by shot number (such as 01-白底主图, the white-background main image), plus a 方案.md with the scene, subject, highlights, copy, shape and every image's prompt. Shapes: portrait 3:4, square 1:1 or landscape 4:3.

How it's made

Describe an idea and add 0–3 reference images (PNG, JPG, WebP), or start from a template: product set, on-model apparel, portrait, event poster, social cover, food spot, home interior.
Click "Let AI plan first": AI looks at the references, reads the brief and drafts a plan with scene, look, subject, highlights, headline and subheading, image count and shape, all editable. Planning is free, up to 30 times per person per 24 hours; you can also skip it and generate directly, or fill in the plan yourself.
Once you confirm, the scene sets out a list of shots, one shot per image, generated one by one; with reference images, the subject's shape and details (for portraits, face and hair) are asked to match the original.
When every image is done they're packed into one ZIP; redo any single image you don't like.

Six scenes and their shots

1–12 images, taking the scene's shots in order; past one round it starts over, and repeated shots change framing and angle.

Product set: true white background (the product cut out of your photo onto pure white, pixels untouched, free), white-background main, selling-point hero, selling point, in use, detail close-up, other angles, mood shot
On-model apparel: full-length front, half-length, side view, fabric detail, back view, selling point, in motion, styled scene. The references are flat-lay or hanger shots of the garment and a model wears it; one model across the set: without a model photo the full-length front is made first and every other shot follows it
Portrait: half-length main, full-length with setting, close-up, candid motion, new scene and light, mood shot
Poster / cover: centered key visual, big-type title, asymmetric layout, graphic illustration
Lifestyle: 45° main, flat lay, eye-level setting, people interacting, another time of day, detail close-up
Creative: main image, cinematic, wide setting, stylized take

What you can choose

References: 0–3 product, garment, portrait or scene photos; on-model apparel can add one model photo so the whole set uses that person
Scene: product set, on-model apparel, portrait, poster / cover, lifestyle, creative
On-model apparel also lets you set the model (gender, age range, hair, build) and the backdrop (plain studio, street, indoor)
Look: auto, minimal, bright, moody, vivid
Images: 1–12; shape: portrait 3:4, square 1:1, landscape 4:3
Subject, highlights (up to 5), headline and subheading, extra notes: all in the plan, all editable

Text on the images

Only selling-point images and posters carry text: the headline, subheading or selling point you wrote in the plan, which the prompt asks to be reproduced exactly. Every other shot has no text, letters, numbers or watermarks.

How you pay

Each image is priced as one image; credits for the whole set are held when you submit, and you see the estimate first. Planning is free. Each image is charged as soon as it succeeds; a failed or rejected image is refunded. A single-image redo is priced separately and charged only if the new image succeeds.

Can I go without reference images?: Yes. A one-line idea is enough to plan and generate; the product set, on-model apparel, portrait and food spot templates need a reference image first.
Is the apparel model the same person in every shot?: Yes. Without a model photo the full-length front shot is made first and every other shot uses it as a reference, asking for the same face, hair and build; with a model photo the whole set uses that person. If the first shot fails the set stops and the other shots aren't charged; adjust and submit again.
Does AI planning cost credits?: No. Each person can plan up to 30 times per 24 hours; after that, fill in the plan yourself and generate directly.
Will the text on the images be wrong?: Only selling-point images and posters carry text, and it's the text you wrote, which the prompt asks to be reproduced exactly. AI can still draw characters wrong; redo that image if it does.
Am I charged for an image that didn't come out?: No. Each image is charged only when it succeeds; credits held for one that failed or was rejected are released.
How is the true white background made?: It isn't generated: a local segmentation model cuts the product out of the photo you uploaded and places it on pure white, so colours and details stay pixel-exact, and it's free. Photos with a clear subject on a simple background work best; if it can't be cut cleanly you're asked for another photo and the rest of the set is unaffected.

One-click images