Answer: Seedream 5.5 generates images for 2 credits at 1K and at 1.5K, and 4 credits at 2K — $0.02 and $0.04 at settlement. It renders text-to-image and image-to-image, edits with up to ten references, splits one image into layers, and tops out at 2K rather than 4K.
That price shape is unusual in a specific way: 1K and 1.5K cost the same. The resolution ladder only steps once, at 2K, which makes 1.5K the quiet default for anything that is not a final asset.
The shape of the model
| Seedream 5.5 | |
|---|---|
| Modes | text to image · image to image · layer decomposition |
| Resolutions | 1K · 1.5K · 2K |
| Credits per image | 2 · 2 · 4 |
| Aspect ratios | auto, 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9, 2:1, 1:2 |
| References | up to 10 (edits), 1 (layer split) |
| Prompt length | 5,000 characters |
| Output | PNG or JPEG · background auto / opaque / transparent |
| Billing | per image; edits and layers settle on usage with a credit ceiling |
When 1K, 1.5K or 2K
1K — 2 credits. Social posts, thumbnails, layout comps, and anything that will be scaled down before it is seen. At two credits an image, exploring ten directions costs 20 credits.
1.5K — 2 credits. The free upgrade: a third more pixels on the long edge for the same price. For web hero images and editorial pages, this is the sensible default, and it is why the 1K tier is mostly for tests.
2K — 4 credits. Double the price, and the right buy when the image is the deliverable: print-adjacent covers, posters, product stills that will be judged at 100%.
There is no 4K. If a spec says 4K, this is not the model — Nano Banana 2.1 carries that tier at 5 credits.
What the three modes actually do
Text to image. A brief in, an image out. The editorial framing is deliberate: the model is pitched at covers, product still lifes and campaign sets rather than at photoreal portraiture.
Image to image. Up to ten references, from your own library, up to 30 MB each in a wide range of formats. This is the edit mode: change the light, restage the object, keep the composition.
Layer decomposition. One source image in, a set of layers out — up to 100, default 30, at a chosen size from 1K to 2K, with optional text recognition. It is the mode that turns a finished render into editable material.
What a month of work costs
| Volume | Setting | Credits | USD |
|---|---|---|---|
| 50 social images | 1K | 100 | $1.00 |
| 50 web heroes | 1.5K | 100 | $1.00 |
| 25 print covers | 2K | 100 | $1.00 |
| 200 drafts | 1K | 400 | $4.00 |
A new account's first week includes 18 credits — nine 1K images, or four 2K images — which is enough to judge whether the look fits.
Where it goes wrong
- Assuming 2K is "just bigger". It is double the price; the 1.5K tier is where most work belongs.
- Buying 4K. It does not exist on this model.
- Treating edits as free. Edits and layer splits settle on usage with a credit ceiling, so the per-image table is not the whole bill.
- Ignoring the ratio set. With eleven ratios including 21:9 and 1:2, composition decisions start before the prompt does.
FAQ
Is 1.5K really the same price as 1K? Yes — 2 credits at both.
Does the free tier include it? Image generation is open to every tier; a new account's weekly credits cover a handful of 1K images.
What is the layer default? Up to 30 layers, configurable between 2 and 100.
Which output format for transparency? PNG, with the background set to transparent.
The model's current price rows are on Callirra's Seedream 5.5 page; read the 1.5K row twice before you default to 2K.