Find Image Generation Models for Your Output Needs

Choose preset aspect ratio and resolution, or send both width and height from 256 to 4096 pixels. The custom ratio must stay between 0.4 and 2.5. Request one to four output images and optionally disable prompt enhancement.

Custom width and height override preset sizing. Total cost depends on pixel tier and image count.

Choose opaque or transparent backgrounds, 1K or 2K resolution, and a supported aspect ratio. Prompt enhancement can be switched off when you want to preserve your original wording.

Transparency requires PNG or WebP. JPEG requires an opaque background.

Use 1K or 2K output. Independent generation accepts one to four images; sequential generation accepts a requested maximum of twelve. For independent outputs, supply a palette of three to ten colors.

Sequential mode disables palette and thinking-mode controls. The model may return fewer images than the requested maximum.

Choose the provider before setting resolution, proportions and background. The SeeAPI route offers background controls at 1K; the Replicate route also documents output format and compression.

On SeeAPI, square output supports 1K and 2K but not 4K. Other resolution combinations also depend on aspect ratio.

The default SeeAPI route returns one image at 1K, 2K or 4K. The fal route supports quality settings and a maximum requested output count of ten. Transparent output on fal requires PNG or WebP.

Do not send fal-only output-count, quality or format fields to the default SeeAPI route. Output count increases the charge.

Use 1K, 2K or 4K on the default SeeAPI route, or choose fal for quality levels, output format and up to four requested images. Keep the prompt and selected provider explicit in comparisons.

Flare and Sunburst have different fal output-count limits. Family membership is not evidence of equivalent image quality.

Choose 1K, 2K or 4K output and PNG or JPEG format. The documented prompt limit is 20,000 characters. Use the aspect ratio that matches your target placement before generating variations.

Higher resolution consumes more credits. A long prompt allowance does not guarantee better instruction following.

The text-to-image documentation lists 1K, 2K and 4K, plus automatic and explicit aspect ratios. Choose the SeeAPI or Replicate provider before building the request.

Check the chosen provider's schema and billing. Do not assume all Nano Banana variants expose identical fields.

The generation endpoint documents 1K, 2K and 4K output, selectable aspect ratios, and PNG or JPEG. Prompts accept up to 10,000 characters.

The public catalog lists a higher access requirement for this model. Check account eligibility before planning an integration.

Basic quality produces 1K output and high produces 2K. Supply a prompt of 3–5000 characters, select one of the documented aspect ratios, and choose PNG or JPEG.

Pro is not the highest-resolution Seedream option in this collection. Account access requirements apply.

Basic, high and ultra correspond to 2K, 3K and 4K output. The endpoint accepts prompts of 3–3000 characters and PNG or JPEG output.

Use an explicit supported aspect ratio. The Pro variant uses a different quality-to-resolution mapping.

Select 1K, 2K or 4K output and a supported square, portrait or landscape aspect ratio. An optional seed can help structure repeated tests.

A fixed seed does not guarantee identical results across provider or model changes.

The SeeAPI route accepts prompts up to 1000 characters and five documented aspect ratios. Replicate adds output format, seed, inference steps and guidance controls.

Replicate-only controls are not available on the default route. Turbo is a model name; no speed benchmark is claimed here.

The SeeAPI route supports prompt upsampling and automatic translation, with prompts up to 2000 characters. Select a documented aspect ratio and JPEG or PNG output.

Provider ratio lists and output-format values differ. Compare upsampling on and off using the same brief.

The SeeAPI route documents prompt upsampling and translation controls. Replicate adds a seed and a different aspect-ratio list.

When SeeAPI automatic translation is disabled, the guide requires an English prompt. No quality advantage over Pro is assumed.

The text-to-image route accepts prompts up to 20,000 characters and automatic or explicit aspect ratios. Use this endpoint's own field list when preparing the request.

The current schema does not expose the resolution selector used by Nano Banana 2 and 2.1.

Generate from a text prompt, select automatic or explicit proportions, and choose the provider's supported output format. The SeeAPI prompt limit is 5000 characters.

Newer family members have different controls. Compare them using the same brief rather than assuming version names establish quality.

The endpoint documents square 1:1, landscape 3:2 and portrait 2:3 output. Prompts accept up to 2000 characters.

Do not copy the resolution, background or output-count controls from GPT Image 2 or 2.5 into this request.

The text-to-image route exposes enable_pro for Quality mode and five aspect ratios: 1:1, 2:3, 3:2, 16:9 and 9:16. Prompts accept up to 5000 characters.

Evaluate both modes on representative briefs. A mode name is not a measured quality or speed ranking.