Concurrent Submit | Zhenzhen Doubao Seedream4.0

Seedream 4.0 is the slightly older, slightly cheaper sibling in the Doubao image line - ByteDance's dependable text-to-image-and-edit model with 1K/2K/4K resolution options and the same character-consistency strengths the family is known for. The blocking node does one image per wait. This submit node is the concurrent version: same doubao-seedream-4-0-250828 call, same controls, but each job drops into the shared 30-worker image pool and returns a task handle instead of a finished image.

The honest framing: if you've already got the 4.5 node working, this one is the budget lane. 4.0 is the model you point at when you want the family's reliability at 1K for fast iteration, or when 4.5 is throttled. Used concurrently it's the cheapest way to carpet-bomb a prompt grid in this pack.

How it works

Auto-generated concurrent wrapper, same machinery as every submit node here: the class copies the base node's inputs at import time, wraps its generation call in a bounded ThreadPoolExecutor (30 image workers by default, set via COMFLY_IMAGE_CONCURRENCY), and returns a COMFLY_IMAGE_FUTURE. Collect with ComflyConcurrent_Image_Await ("Concurrent Collect Images (30)"), which waits for the batch and returns IMAGE tensors in slot order.

Required inputs: prompt, model (default doubao-seedream-4-0-250828), response_format (url default - lighter than base64 at batch scale), and resolution (1K / 2K / 4K). Optional: aspect_ratio (ten options incl. Custom with width/height), apikey, image1-5 (references for editing), sequential_image_generation, max_images (up to 15), seed (default -1), watermark (on by default - the flip side of the 4.5 node's default), stream, and skip_error.

Installing it

One clone, whole pack:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

Restart ComfyUI, or install "Comfyui-zhenzhen" via ComfyUI Manager. No model downloads - Seedream runs on ByteDance hardware behind the proxy.

Common issues

Pack-level: account + credit required, 443 errors need VPN with TUN mode, first-run 500s are usually upstream - re-run. Seedream-specific: the watermark default differs from 4.5 - it's on here, so un-stamp your batch if the output's watermarked. And the cost ladder is real: 30 concurrent 4K calls is the most expensive thing in this article. The sane batch workflow is 1K validation first, 2K for near-keepers, 4K only for the finalists.