Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Seedream 4.5 & 5.0

Node name
BytePlus Seedream 4.5 & 5.0
Node ID
BytePlusSeedream
Find it
Double-click the canvas and search for “Seedream 4.5 & 5.0”, or browse BytePlus ModelArk
Keys needed
ModelArk API key
Runs when
Its output is used by another node, such as Save Image or Save Video. It is not an output node.

What it does

Section titled “What it does”

Creates images with BytePlus Seedream from a text prompt. Connect reference images to edit a picture, combine several, or keep a subject consistent: up to 10 on Seedream 5.0 Pro and Flash, and up to 14 on 5.0 Lite, 4.5 and 4.0. Each model option shows its own inputs, so pick the model first.

The node bills for every image it generates. It does not save anything itself: connect Save Image to keep the result.

On the canvas

Section titled “On the canvas”
The Seedream node with the model set to seedream 5.0 pro, a prompt about a glossy 3D paper crane badge, the size preset (1K) 1024x1024 (1:1), and the advanced settings below it
Seedream 5.0 Pro selected, with one free image slot. The inputs below the model change with the model.

Inputs

Section titled “Inputs”
InputTypeDefaultWhat it does
prompttextText prompt for creating or editing an image.

model is required and decides which of the inputs below appear.

InputTypeDefaultWhat it does
size_presetchoicePick a recommended size. Select Custom to use the width and height below.Choices: (1K) 1024x1024 (1:1), (1K) 864x1152 (3:4), (1K) 1152x864 (4:3), (1K) 1312x736 (16:9), (1K) 736x1312 (9:16), (1K) 832x1248 (2:3), (1K) 1248x832 (3:2), (1K) 1568x672 (21:9), (1.5K) 1536x1536 (1:1), (1.5K) 1344x1792 (3:4), (1.5K) 1792x1344 (4:3), (1.5K) 2048x1152 (16:9) … (28 in total)
widthinteger2048Custom width for image. Value is working only if `size_preset` is set to `Custom`Range: 240–8600
heightinteger2048Custom height for image. Value is working only if `size_preset` is set to `Custom`Range: 240–8600
imagesgroup of inputs of IMAGEOptional reference image(s) for image-to-image or multi-reference generation. Up to 10 images.Slots: image_1 to image_10 (10)
seedinteger42Seed to use for generation.Range: 0–2147483647
reference_maskoptionalMASKAlpha of the reference image for a transparent background (connect the MASK output of Load Image).
Advanced inputs for seedream 5.0 pro
InputTypeDefaultWhat it does
prompt_optimizationchoicestandardPrompt-optimization mode when reference images are provided: 'standard' gives higher quality, 'fast' shorter generation time.Choices: standard, fast
watermarkon/offoffWhether to add an "AI generated" watermark to the image.
thinkingon/offonEnable the model's prompt-optimization reasoning ('thinking') for better adherence. Can substantially increase generation time — notably on Seedream 5.0 Pro. Can only be disabled for text-to-image (not when reference images are provided).
generation_countinteger1Number of separate generations to run in parallel. With several, generation N uses seed + N so the results differ.Range: 1–16
output_formatchoicejpegFile format of the generated image. A transparent background needs png.Choices: jpeg, png
backgroundchoiceopaquetransparent: edit one reference image that has an alpha channel (connect its mask to reference_mask) and return a PNG with transparency, also as the mask output. Needs output_format png.Choices: opaque, transparent

Notes on inputs

Section titled “Notes on inputs”
  • Size: size_preset lists ready-made sizes for 1:1, 3:4, 4:3, 16:9, 9:16, 2:3, 3:2 and 21:9. The entries ending in “(adaptive)” send only a resolution level and let the model choose the ratio. Choose Custom to use width and height. The size limits are under Limits and tips.
  • Reference images: connect them to the images slots (image_1, image_2, …) in order. With reference images the model edits or combines them following the prompt.
  • Related image sets (5.0 Lite, 4.5 and 4.0): set max_images above 1 to let one request return between 1 and max_images related images, for example story scenes or character variations. Reference images plus max_images cannot exceed 15. With fail_on_partial on, the run aborts if any requested image is missing or returns an error.
  • Transparent background (5.0 Pro and Flash): connect exactly one reference image, set output_format to png, set background to transparent, and connect the MASK of Load Image to reference_mask.
  • Several images in one run: raise generation_count. The runs go in parallel and generation N uses seed + N, so the pictures differ.

Outputs

Section titled “Outputs”
OutputTypeWhat it holds
IMAGEIMAGE
responsetextRequest metadata of each generation as JSON.
maskMASKTransparency of the output (1 = transparent, like Load Image). Empty unless background is transparent.

Notes on outputs

Section titled “Notes on outputs”
  • With generation_count or max_images above 1, IMAGE holds all pictures as one batch. A picture of a different size is resized to the size of the first one.
  • mask is fully opaque (all zeros) unless background is transparent. Then it holds the transparency of the result, with 1 meaning transparent, like Load Image.

Limits and tips

Section titled “Limits and tips”
  • Activate each Seedream model in the ModelArk console for the region you use. Seedream 5.0 Lite is not available in eu-west-1: the node stops before it sends anything.
  • The size is checked before anything is sent. The ratio must be between 1:16 and 16:1, and the total pixels must fit the model: 921,600 to 4,624,220 on 5.0 Pro and Flash, 3,686,400 to 16,777,216 on 5.0 Lite and 4.5, and 921,600 to 16,777,216 on 4.0.
  • On 5.0 Pro the 1.5K sizes are priced like 1K.
  • thinking is on by default (5.0 Flash has no such setting). It can substantially increase generation time, most on 5.0 Pro, and it can only be switched off when no reference image is connected.
  • For reference images of a character, a place or an object, the Production assets templates (Character Sheet, Location Sheet, Prop Sheet) fill a fixed prompt from a few text boxes and send it to this node.
  • A rejected result (“Generated image contains sensitive content”) can be a false alarm. Run again with another seed or reword the prompt.
  • If you interrupt the run, the node stops waiting. A request that was already sent may still be billed.
  • A failed request that may have billed is never sent again automatically. See Billing and cost.

Common errors

Section titled “Common errors”
Message What to do
No BytePlus API key is set Add the key in Settings → BytePlus. See API keys.
has not activated the model Open the ModelArk console and activate the model for the same region as your key.
is not available in The model is not offered in your region, for example Seedream 5.0 Lite in eu-west-1. Pick another model, or another region in Settings → BytePlus.
Generated image contains sensitive content Run again with another seed or rephrase the prompt.
Rate Limit Exceeded (429) The node already waited and retried twice. Wait a moment, or lower generation_count.

Used in templates

Section titled “Used in templates”