Create cinematic AI images from a prompt

Write the shot, pick a model, choose shape and size, then generate. A step-by-step guide to Create image with prompt structure, model choices and live prices.

BeginnerTutorial video 0:262 min read

Create image turns a written description into a picture. Write it like a director describes a shot — subject, setting, light, lens — and you get stills that work as finished images or as first frames for a video.

Close portrait of a weathered fisherman in a yellow oilskin in the rain at dusk
A portrait made from one prompt with GPT Image 2.5 Sunburst.

Step 1: Open Create image

Open Create image. The prompt bar sits at the bottom with the model, style, shape and number of images next to it.

The Create image page with example images on top and the prompt bar with model, style and Generate below
The model shown in the title is the current default; switch it from the prompt bar.

Step 2: Write the shot

Start with the subject, then the setting, then light and camera. Concrete details beat adjectives: "sodium harbor lights" says more than "moody".

The prompt bar filled with a description of a fisherman portrait at dusk
Subject, setting, light, lens — in that order.
Try this promptModel: Seedream 5.0 Flash

Weathered North Sea fisherman in a yellow oilskin, close portrait at dusk, rain on his beard, sodium harbor lights behind him, shallow depth of field, shot on 85mm film, natural skin texture.

Step 3: Pick shape and size

Choose an aspect ratio for where the image will live: 3:4 or 4:5 for portraits and feeds, 16:9 for banners and video first frames, 9:16 for stories. Pick a size; Standard is enough to judge the idea.

The aspect ratio picker open with square, portrait, landscape and tall options
Pick the shape before you generate; changing it later means generating again.

Step 4: Choose how many and generate

Generate up to four images at once to compare options. Each image is charged, and the Generate button shows the total. With the default Seedream 5.0 Flash, one image costs 29 credits.

The number of images set to 2 next to the Generate button showing its price
Two to four images is a good way to explore before you commit.

Tips for better results

  • Use Write it for me to turn a rough idea into a detailed prompt (it costs a few credits).
  • Add a reference image when you need a specific face, product or style to carry over.
  • Avoid text in the image unless the model is known for it; add type later in a design tool.
  • Next step: send a still to Create video to make it move.

FAQ

Which image model is best?

It depends on the job. Seedream 5.0 Flash is fast and cheap for drafts; Nano Banana and GPT Image are strong at following detailed instructions and editing. Compare them on the models page.

Can I edit an image I already have?

Yes. Switch the mode to Edit image, add your picture and describe the change.

Can I use the images commercially?

You can use what you make for your own projects, subject to the terms. Do not upload photos of people without their permission.

Next tutorials

All tutorials

Upscale2 min read

Upscale an image to 4K or 8K

Add a photo, choose Enhance or Upscale only, pick the content type and size, then generate. How to upscale images to 2K, 4K or 8K and what each option costs.

Try it while it’s fresh

The tool opens with the settings from this tutorial. You’ll see the credit price on the button before anything runs.