Guides

Seedream images

Generate images from a prompt, or edit and blend up to 10 reference images, in chat or through the API.

seedream-pro is the premium image model. Give it a prompt and it generates an image; give it one or more reference images and it edits, transforms, or blends them, holding your subject consistent. It is the model to reach for when you want to change an existing image, combine several into one, or keep a character looking the same across generations.

Start from a strong input

For a plain generation, a detailed prompt is enough. For an edit, the reference images do most of the work: the model keeps their subject, composition, and style, and applies what your prompt asks for. A clear, high-resolution reference gives a cleaner result than a small or noisy one.

Three ways to make an image

The mode is chosen automatically by how many reference images you attach:

  • No references (generate). Describe the whole image. Best for making something new from scratch.
  • One reference (edit). Transform a single image: change a background, restyle it, add or remove an element. Describe the change, not the whole scene.
  • Two or more references, up to 10 (blend and subject). The images are ingredients: blend them, transfer a style from one onto another, or keep a subject consistent. Cite them in the prompt to say what each one is for.

Citing references with @image

When you attach more than one reference, refer to each in the prompt by its number: @image1, @image2, and so on, in the order you added them.

"put the character from @image1 into the scene from @image2, keep the lighting of @image2"

In chat, the reference builder numbers each thumbnail; tap a slot to drop its @image token into the prompt at your cursor. In the API, send the images as image and cite them the same way. Order matters: the first image you attach is @image1.

In chat

Select Seedream Pro in the composer and a reference builder appears above it:

  • Add references with the tile, up to 10. Each becomes a numbered slot.
  • Tap a slot to cite it (@image1, and so on) in your prompt.
  • Set the size in the toolbar. The price updates live as you change it.

In the API

Seedream editing runs on the OpenAI-compatible /v1/images/edits endpoint. Pass model: seedream-pro, your prompt, and send references as repeated multipart image fields, citing them as @image1, @image2 in the prompt:

curl https://api.roteo.ai/v1/images/edits \
-H "x-api-key: $ROTEO_API_KEY" \
-F model=seedream-pro \
-F "prompt=put the character from @image1 into the scene from @image2" \
-F size=2K \
-F image=@character.jpg \
-F image=@scene.jpg
bash · 7 lines

To generate without references, use /v1/images/generations with the same model. See Images for the full request.

Size and cost

  • Size: 1K, 2K, or 4K.
  • Cost is per delivered image and scales with size. Editing adds a small per-reference charge: the first reference is free and each additional one costs a little more, up to the limit of 10. A failed generation is free. Live pricing is in the dashboard.

Writing good prompts

  • Detail is free. The bill depends on the image's size and how many references you attach, never on how long the prompt is.
  • For an edit, describe the change, not the whole scene. The reference already fixes the look, so say what is different.
  • When blending, name each reference. Cite @image1, @image2 and spell out what each contributes (subject, style, background), so the model knows how to combine them.
  • Use clear reference images. A sharp, well-lit photo edits cleanly; a tiny thumbnail or heavily compressed image gives the model less to work with.
Seedream images | Roteo