Concurrent Submit | Zhenzhen Doubao Seedream3.0

Seedream 3.0 is where the Doubao image line started in this pack - the t2i model that predates the 4.0/4.5 refinements and still works fine for straightforward generation. The blocking node does one image per call. This submit node is the concurrent version: same doubao-seedream-3-0-t2i-250415 call, same size and guidance controls, but each job goes into the shared 30-worker image pool and returns a task handle instead of a finished image.

Reach for it when you want the cheapest possible batch lane in the Zhenzhen Doubao family. 3.0 is slower and less refined than its successors, but for draft work - "I need twelve rough concepts to pick three from" - the concurrent version makes the speed difference irrelevant, because twelve jobs run at once. It's the sensible first rung of the ladder: validate on 3.0 at 30-way, then re-roll the survivors on 4.5.

How it works

Auto-generated concurrent wrapper, identical machinery to 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-3-0-t2i-250415), response_format (url default), size (nine presets from 1024x1024 up to 1512x648, plus Custom where the Custom_size string like "1536x1024" takes over), and guidance_scale (2.5 default). Optional: apikey, seed (default -1), watermark (on by default), 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-3.0-specific: Custom_size is a freeform string, not a dropdown - get the WxH format wrong (e.g. spaces, lowercase x) and the request can fail at the API rather than the node. And remember 3.0 is the budget tier for a reason; if a batch comes back looking dated next to 4.5, that's the model, not your prompt - re-roll keepers on the newer node rather than re-prompting this one. Meter note: 30 concurrent calls, 30 paid calls - the pool is parallelism, not discount.