Image generation

Turn a text prompt, or your own reference photos, into images with 37 models.

Overview

The image studio at /generate-image has three parts: the prompt and settings on the left, the canvas in the middle, and your recent creations in the filmstrip and history panel.

Write a good prompt

Describe the subject, the setting, the lighting and the style, in that order.
Add camera words for photos: 35mm, shallow depth of field, golden hour.
Use the Negative Prompt for things to avoid, for example blurry, distorted, extra fingers.
Pick an Aesthetic Preset to add a tested style without writing it yourself.

Portrait of an elderly fisherman mending a net on a wooden pier,
soft morning fog, warm rim light, 85mm lens, shallow depth of field,
photorealistic, muted colour grade

Settings

Setting	What it does
Model	Which AI model renders the image. Each shows its price and strengths.
Aspect Ratio	The shape of the image: 1:1, 4:5, 9:16, 16:9 and more.
Aesthetic Preset	A ready-made style (cinematic, anime, product shot and more).
Guidance Scale (CFG)	How strictly the model follows your prompt. Higher is stricter but can look harsher.
Sampling Steps	More steps add detail but take longer, on models that support it.
Seed	`-1` is random. Reuse a seed to get a similar result again.

Reference photos

Models such as Nano Banana, FLUX.2, Flux Kontext and Seedream accept reference images. Add up to 4 or 10 photos, depending on the model, and describe the change you want, for example put this product on a marble table.

Photo edits are priced separately from text-to-image. The picker shows the edit price before you generate.

After generating

Variations makes new versions of the same idea.
Upscale and Remove BG are covered in Edit and enhance.
Motion animates the image into a video. See Video generation.
Public / Private controls whether the image appears on Explore. New images are public by default.
Download Master saves the full-quality original.