Image mode turns a written prompt into a 2D image you can use as concept art, a texture reference, or the starting point for a 3D model.
Generate an image
- Switch the workspace to the Image tool.
- Write a clear, descriptive prompt (up to 1,500 characters) — subject, style, lighting, and mood all help.
- Choose an AI model and options (ratio, style, count, and more) — covered below.
- Press Generate. The button shows the credit cost of the job.
The result usually appears in your library within a few seconds.
Choosing an AI model
You can pick from several image models, each with its own look, strengths, and credit cost.
- Nano Banana — the default model, available on the Free plan.
- Nano Banana 2 · Nano Banana Pro — higher-fidelity Google models.
- GPT Image 1 mini · GPT Image 1.5 · GPT Image 2 — the OpenAI lineup.
Generation options
Below the model you can adjust:
- Count — generate 1–4 images at once. More images cost proportionally more credits.
- Aspect ratio — choose 1:1, 3:2, 2:3, 16:9, or 9:16. (OpenAI GPT Image models support 1:1, 3:2, and 2:3 only.)
- Style preset — None, Japanese Anime, Chibi Blind Box, Pixel Art, Concept Art, or Isometric Low-poly. The chosen style is woven into your prompt automatically.
- Reference image — attach one image to steer color or composition.
- T-pose — generate a character in an arms-out T-pose, which is ideal for characters you'll later use in Animation.
- Pure white background — produce a clean white background, ideal for sending into Image to 3D.
4-view (multi-view) sheets
Turn on multi-view to generate one character seen from four directions — front, back, left, and right — in a single sheet, perfect as input for multi-view 3D generation. Multi-view locks the ratio to 1:1 and the count to 4, and auto-enables the white background (so it costs four images' worth of credits).
Writing a good prompt
- Lead with the subject, then add style and details ("a low-poly fox, flat colors, soft studio lighting").
- Mention the look you want for games: isometric, flat-shaded, hand-painted, pixel art. The style presets help with this.
- Keep one idea per prompt — raise the count to get variations rather than cramming everything into one line.
Next steps
From a generated image you can tweak the prompt and generate again, or send it straight into Image to 3D to turn it into a model. Refining results is covered in Editing & Re-generating.