Seedream 5.5 Edit

ByteDance Seedream 5.5 editing with up to 10 reference images: change one thing and keep every unedited pixel, round after round, or pull the subject out as its own transparent layer

Last updated View as MarkdownAgent setup

Seedream 5.5 Edit

Node type: edit:bytedance_seedream_v5_5_edit
Category: edit
Hidden in sidebar: yes (auto-selected by routing)

Description

ByteDance Seedream 5.5 editing with up to 10 reference images: change one thing and keep every unedited pixel, round after round, or pull the subject out as its own transparent layer

Pricing

  • Cost: ~13 credits per image

Variant family

  • Family: bytedance_seedream_v5_5
  • Is primary: no
  • All variants: generate:bytedance_seedream_v5_5, edit:bytedance_seedream_v5_5_edit

Canvas ports

These appear as port handles on the left side of the node.

ID Label Details
prompt Prompt TEXT (required) · multiline
image1 Image 1 IMAGE

These render as form fields in the right-side config panel when the node is selected.

ID Label Details
size Size TEXT · options: 2K, 1.5K, 1K, 2048x2048, 2816x1584, 1584x2816, 2368x1776, 1776x2368, 2496x1664, 1664x2496, 3136x1344 · default: "2K"
num_images Number of Images NUMBER · 1…4 · default: 1
output_format Output Format TEXT · options: png, jpeg · default: "png"
background Background TEXT · options: auto, transparent, opaque · default: "auto". Transparent returns the subject on a see-through background (a PNG with an alpha channel), ready to use as a sticker or a layer. Needs PNG output.

Outputs

ID Label Type
images Generated Images IMAGE

How to prompt Seedream 5.5

  • Describe the photo, not the idea. Name the subject, the setting, the light and the lens: "a ceramic mug on a walnut desk, soft window light from the left, 50mm, shallow depth of field". Lighting, materials and skin are where 5.5 improved most, so say what they should look like.
  • Text in the image. Put the exact words in quotes and say where they go: "a cafe chalkboard that reads 'OPEN 7AM'". Small text renders better than in 5.0, but keep it short.
  • Transparent cutouts. Set Background to Transparent (PNG output) and describe only the subject: "a red sneaker, side view". The result is a PNG with an alpha channel, ready to drop into the Compositor or the Video Editor as a layer. It works with reference images too, and they do not need an alpha channel of their own. Opaque forces a solid background; Auto lets the prompt decide.
  • Edits. Wire the image and give one change per run: "change her jacket to green. Keep everything else the same." Unedited areas stay pixel for pixel, so chain Edit nodes to make one change at a time instead of one long instruction.
  • Reference images. Up to 10. Type # in the prompt to insert "Figure 1", "Figure 2" and so on, in the order they are wired, and say what each one is for: "the woman from Figure 1 wearing the jacket from Figure 2".
  • Size. Auto 2K, Auto 1.5K and Auto 1K pick the shape from your prompt and images, so say the ratio in the prompt ("a 16:9 banner"). Choose a fixed 2K size (16:9, 9:16, 4:3 and so on) when the layout must be exact. 1K and 1.5K cost less per image than 2K.
  • Credits. Each image is priced by its size, plus a little for every reference image wired in. The cost on the node already counts both.

Auto-generated from the Wireflow node registry.

For AI agents: the full documentation index is at /llms.txt, and most docs pages are available as Markdown by adding .md to the URL.

© 2026 Wireflow. All rights reserved.