Skip to content

Model & resolution ​

The sidebar and Direct Edit share image models and settings. Model, resolution and quality determine the base credit price per image. See Credits & rules.

Image models ​

ModelSuggested useResolutions
Seedream 5.0 ProStrong all-around editing · Slower generation1K, 1.5K, 2K
Seedream 5.0 FlashStrong editing · Faster generation1K, 1.5K, 2K
Nano Banana ProHigh-quality general generation1K, 2K, 4K
gpt-image-2High-fidelity complex generation1K, 2K
Nano Banana 2Fast and reliable · Everyday choice512, 1K, 2K, 4K
gpt-image-2.5-flareFast generation · Efficient creation1K, 2K
gpt-image-2.5-sunburstPrecise editing · Composition adjustments1K, 2K

New tasks default to Nano Banana 2 + 1K + Auto aspect ratio. History still restores the saved model and settings, including models hidden from new-task selection.

Seedream 5.0 Pro ​

Seedream 5.0 Pro supports text-to-image, single-image and multi-reference generation, plus Global and Region Direct Edit. The product currently exposes 1K, 1.5K, and 2K only; 4K is not available.

Seedream Auto aspect-ratio behavior:

  • Text-to-image: 1:1;
  • Image-to-image: the real aspect ratio of the first effective sidebar input (Image 1). If Image 1 is empty while Image 2 or Image 3 has an image, the product does not silently use Image 2; move the main reference to Image 1 or choose a ratio explicitly;
  • Direct Edit: the real aspect ratio of the image currently being edited; an explicitly selected preset takes priority.

1K and 1.5K use the same user-credit tier, while 2K uses the higher tier. See Credits & rules.

Seedream 5.0 Flash ​

Seedream 5.0 Flash supports text-to-image, single-image and multi-reference generation, plus Global and Region Direct Edit. It offers 1K, 1.5K, and 2K at 5 credits per image for each resolution; 4K is not available. On first selection without saved Flash settings, the default resolution is 2K. Auto and preset aspect ratios follow the same rules as Seedream 5.0 Pro. See Credits & rules for details.

GPT Image quality and aspect ratio ​

GPT Image 2, GPT Image 2.5 Flare, and Sunburst use model-specific base credits. The current open range is 1K / 2K with Low / Medium quality. Defaults are 1K, Low, Auto, and the estimated cost is shown before generation. See Credits & rules.

Model1K Low1K Medium2K Low2K Medium
gpt-image-210201525
gpt-image-2.5-flare8121020
gpt-image-2.5-sunburst8121020

This GPT Image pricing update covers the 1K / 2K Low / Medium tiers. High and 4K are not in the current open range.

Existing smart-template and prompt-enhancement fees apply under their respective rules. LightReference is a separate lighting-editing feature; it does not use the sidebar GPT quality setting or the GPT Image quality matrix.

Available ratios are Auto, 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9 and 21:9. With input images, Auto follows the aspect ratio of the first image uploaded in the sidebar (the first image sent to the model); text-only generation defaults to a square image.

Switching models ​

Settings adjust to the new model's supported resolutions and ratios. Unsupported extreme ratios reset to Auto. Rerunning an image from History restores its saved model, quality, resolution and ratio.

For UI or rule updates, please refer to the actual product.