Text to Image

Pick an AI model, style, and aspect ratio, then turn a prompt into concept art, reference images, or images ready to become 3D.

The Image tool turns a written prompt into a 2D image. Use it as concept art or a reference, or send it straight on as the starting point for a 3D model.

Generate an image

  1. Choose Image in the tool bar on the left.
  2. Pick an AI model — covered below.
  3. Write your prompt (up to 1,500 characters). Any language works.
  4. Optionally set a reference image, style, aspect ratio, and number of images.
  5. Press Generate. The credit cost of the job is shown just above the button.

Results land in the image feed in the middle of the screen. Depending on the model, that takes anywhere from a few seconds to about two minutes. If a generation fails, its credits are refunded automatically and the card offers Retry.

Choosing an AI model

The model picker groups models by what they're for. Each model's credit cost is shown right in the picker.

High Quality

  • GPT Image 2.5 Flare — higher quality than GPT Image 2, and faster.
  • GPT Image 2.5 Sunburst — the top-precision model, built for polished, detailed edits.
  • Nano Banana Pro — Google's highest-quality model.
  • GPT Image 2 (available on Free) — premium quality with a distinctive style.
  • Grok Pro — xAI's premium model, with refined detail.

Balanced

  • Nano Banana 2 — Google's model with improved speed and quality.
  • GPT Image 1.5 — fast, with balanced quality.
  • Grok Quality — xAI's balance of speed and quality.

Fast · Economical

  • Nano Banana (available on Free · selected by default) — the fast, efficient default.
  • Grok — fast and affordable.
  • PB Image (available on Free) — PicoBerry's fast, lightweight image model.

Tools in the prompt box

Below the prompt box sit a few chips that shape the result:

  • 3D — composes a single subject, centered on a white background. Turn it on if you plan to turn the image into 3D. (It has no effect when a reference image is attached or when you're iterating.)
  • T-Pose — draws a humanoid character in an arms-out T-pose, ideal for characters you'll later animate.
  • 4-View — makes four images of the same subject from four directions in one go (see below).
  • ✨ Prompt Magic — lets AI write or polish your prompt (sign-in required, no credits).
    • Generate Detailed Prompt — expands what you've written into a detailed prompt. With an empty box, it writes a fresh example prompt.
    • Enhance with AI — describe the change you want ("make it red", "add dramatic lighting") and it rewrites the prompt.

Generation options

  • Reference image — attach one image with Add reference images to steer style and composition. Upload one (PNG, JPG, or WebP, up to 20 MB) or pick from Your Generations or Collections.
  • Stylization — None, Japanese Anime, Chibi Blind Box, Pixel Art, Concept Art, Isometric Lowpoly. The style is worked into your prompt automatically.
  • Aspect ratio — the Nano Banana and Grok families support 1:1, 3:2, 2:3, 16:9, and 9:16; the GPT Image family and PB Image support 1:1, 3:2, and 2:3. Switch to a model that doesn't support your current ratio and it snaps to the nearest supported one.
  • Number of images — 1 to 4 per run, each costing credits. How many you can run at once follows your plan's concurrent-generation limit; the Free plan makes one image at a time (4-View excepted).

Generated images are always saved as private — they're never published to Explore.

4-View — four directions at once

Turn on 4-View to get four images of one subject — front, right, back, and left — in a single run. The front view is drawn first and the other directions are drawn to match it, so all four show the same subject.

  • The image count is fixed at 4, and so is the cost (four images' worth).
  • The 3D chip turns on with it.
  • It isn't available on Nano Banana Pro (the chip turns grey).

A finished 4-View set gets a Send to Multi-view 3D button. Choose → 3D Mesh to drop all four into the multi-view input of Image to 3D, or → 3D Texture to use them as the multi-view input of Texture.

Monthly Nano Banana allowance (paid plans)

On a paid plan you also get a separate monthly allowance of credits that only work for Nano Banana image generation (an open event running through December 2026). It's spent before your regular credits whenever you generate with Nano Banana, Nano Banana 2, or Nano Banana Pro, and when the allowance covers a job you'll see Charged to your Nano Banana allowance · N left under the cost. The amount depends on your plan and doesn't roll over to the next month. See Credits & Plans.

Writing good prompts

  • Lead with the subject, then add style and details ("low-poly fox, flat colors, soft studio lighting").
  • Name the look you want: isometric, flat-shaded, hand-painted, pixel art. The style presets help with this.
  • Keep one idea per prompt, and get variations by raising the image count or iterating.

Next steps

Refine your images in Editing & Re-generating, or send one straight to Image to 3D to turn it into a model.