Nodes/BizyAirPlus/ByteDance Seedream V4 Edit (Official Edition)
ComfyUI Node

ByteDance Seedream V4 Edit (Official Edition)

Image-to-image through the official API

By siliconflow·Created 5 months ago·Updated about a month ago· 20
ByteDance Seedream V4 Edit (Official Edition)
  • images
  • api_config
  • image
  • request_id
  • urls
prompt
size2048x2048
skip_errorfalse

The image-to-image sibling of Seedream V4: hand it one or more images and a prompt, and it returns an edited image that respects both. It runs in the cloud through BizyAirPlus's Official Edition route - no local model, no GPU.

Seedream V4 Edit is what you reach for when text-to-image gives you 90% of the way there and you need to push the rest: change the background, swap the lighting, restyle a subject, fix a detail. Because the edit runs through ByteDance's API, it benefits from a model that's much better at instruction-following than typical open checkpoints, which matters a lot in this use case - image editing is where a model that ignores half your prompt becomes obvious. It's also the cleanest way to iterate: regenerate with a tweaked prompt instead of re-rolling from scratch.

How it works

Output node in the BizyAirPlus cloud layer. Flip BizyAirPlus ON in the action bar, wire your reference images into the images input (a COMFY_AUTOGROW_V3 socket, so it accepts single images or lists), and the node forwards them plus your prompt to Seedream's edit endpoint. The edited image comes back as a normal IMAGE tensor. Official Edition just names the API route - the vendor's direct one rather than BizyAir's channel.

The inputs that matter

  • images - the reference(s) to edit. This is the whole point of the node; feed it the image you want changed.
  • prompt - what to change. Be specific about the edit ("make it sunset lighting," "replace the background with a city street") - Seedream rewards describing the edit, not re-describing the whole image.
  • size - output dimensions as a string, default 2048x2048.

Then the two helpers that appear on every BizyTRD node: api_config (per-request base URL/API key override) and skip_error (placeholder output on failure instead of a workflow abort).

The outputs

image wires downstream like any image - save it, upscale it, chain another edit. urls is the direct link, request_id the trace ID for support.

Installing it

Ships in BizyAirPlus - ComfyUI Manager search "BizyAirPlus", or:

cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart, turn BizyAirPlus ON, and enter the key from bizyair.ai (stored at ~/.BizyAirPlus/apikey.ini, or set BIZYAIR_API_KEY). The node logic is in the bizytrd / bizyair-cloudberry pip packages; no model downloads, no VRAM requirements.

Gotchas

Same first-run traps as every cloud node: button OFF or missing key, with the real error in the ComfyUI console. And the standing reputational note - BizyAir is legit SiliconFlow software, not the malware some old threads claimed; that was a Manager auto-install mixup, fixed. For the node itself, the size field is a plain string, so 2048x2048 works and spacing variants may not. When an edit keeps drifting from your intent, simplify the prompt to name the single change you want rather than stacking five - that's where Seedream's edit models are strongest.

CategoryBizyAirModelServices/BizyTRD/Seedream

Inputs (5)

NameTypeDefaultDescription
imagesCOMFY_AUTOGROW_V3
promptSTRINGPrompts
sizeoptSTRING2048x2048Specifies the width and height pixel values of the generated image
api_configoptBIZYAIR_OPENAPI_CONFIG单次请求覆盖base_url和api_key
skip_erroroptBOOLEANfalse开启后遇到错误不中断工作流,输出对应类型的错误占位符

Outputs (3)

NameTypeDescription
imageIMAGE
request_idSTRING
urlsSTRING