Build the scene, pick the characters and send your request: the images are created on the server and arrive here, in your browser.
Sign in with Google: you can generate 20 images a day, one at a time.
When you ask for several images in one request (No. field) you can make part of the prompt change on every image: the 1st uses the first variant, the 2nd the second, and so on; when the variants run out it starts again. Nothing changes if the prompt has none of these symbols.
wearing a [red|blue|yellow] dress: the brackets pick one word inside the sentence (several groups are fine).(red|blue:1.2): with a weight, which stays on every variant → (red:1.2), then (blue:1.2). ([red|blue]:1.2) works too.from side | from above | from behind: a comma-separated piece with the variants split by |.Several groups in the same prompt rotate together (same image number): with a different number of entries the combinations shift on every round. Brackets without | stay normal, and one group cannot be nested inside another.
*name*Written with asterisks, they are replaced by a ready-made sequence, one entry per image. At the start of a series the order is shuffled at random and then followed, so no entry repeats until all have been used. You can mix them with alternatives (which follow the order you write).
*turn* | from left, from front, from behind, from right, from above, from below |
*shoting* | close-up, portrait, upper body, cowboy shot, full body |
*expr* | smile, laughing, blush, angry, crying, surprised |
*pose* | standing, sitting, kneeling, lying, leaning forward, walking |
*angle* | dutch angle, wide shot, fisheye, close-up, from above, from below |
*light* | sunlight, backlighting, night, sunset, rim light, soft lighting |
*time* | morning, noon, sunset, twilight, night, rainy day |
*season* | spring, summer, autumn, winter |
*hair* | blonde, black, red, blue, silver, pink hair |
*outfit* | school uniform, dress, casual clothes, kimono, hoodie, jacket, sailor uniform, business suit, sweater, yukata, maid outfit, winter coat |
*cosplay* | witch costume, pirate costume, vampire costume, knight armor, ninja outfit, princess dress, magical girl outfit, samurai armor, steampunk outfit, fairy costume, astronaut suit, superhero costume |
*gaze* | looking at viewer, looking away, looking back, looking down, looking up, closed eyes |
*style* | watercolor, sketch, flat color, oil painting, pixel art, lineart |
Example: 1girl, *shoting*, *turn*, smile with No. = 6. A name that does not exist is left as it is. Note: "✨ Optimize with AI" may rewrite or drop these symbols, so use them on the final prompt. The number of images per request is limited per account.
Roughly draw shapes and colors (sky, figure, floor…): the model turns them into an image following the prompt. For a sketch a high strength works best, 0.85-0.95: lower stays a drawing.
Paint the part to redo (a bit larger is fine), write in the prompt what you want there and press Retouch: the rest of the image stays identical. It arrives in "Your images".
Drag the dots to bend arms, legs and head; drag a line to move the whole person. Each figure is a person: still describe them in the prompt (who they are, what they wear, where they are). Strength 1 = follows the pose; lower makes the prompt count more.