并发提交｜Zhenzhen Image GK v1.5 图像生成/编辑

The Zhenzhen Image GK v1.5 concurrent submit node is a two-model box: zhenzhen-image-gk-v15 for text-to-image and zhenzhen-image-gk-v15-edit for image editing, switched by the model dropdown. GK is the shop's mid-range image family (G is the budget line, GK is a step up, and if you squint the naming ladder reads like the reseller's own tier list). This node's party trick for a batch workflow is that it's one of the smallest input surfaces in the pack - five controls and a single reference slot - so wiring up a dozen parallel calls takes about a minute.

Inputs:

• model - zhenzhen-image-gk-v15 or zhenzhen-image-gk-v15-edit.
• prompt - required, up to 20,000 characters, for both models.
• size - 1:1, 16:9, 9:16, 3:2, or 2:3. No custom option here.
• n - 1 to 10 images per request; the node downloads the gateway's primary result.
• image1 - required by the edit model, ignored by the text model. And here's the honest quirk: only image1 is submitted. The edit model takes exactly the first reference image, no more. Don't go hunting for image2 on this one.

Output is the single future (SEEDANCE_IMAGE_FUTURE), into 并发接收图片（30 路）.

How it works

Same as every 并发提交｜ node: inputs validated synchronously against the selected model, the job handed to the image thread pool (up to 30 workers; SEEDANCE_IMAGE_CONCURRENCY to dial down), future returned immediately. The receiver collects all slots in parallel and hands back images in order, with per-slot skip_error and the receiver's failure_mode governing failure handling. Mix it freely with other image futures - the receiver only cares that each slot is an image future.

Installing it

ComfyUI Manager → search ComfyUI Seedance, or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
python -m pip install -r ComfyUI_Seedance/requirements.txt

Restart, look under Seedance/并发提交. The dependency is requests; the key is from api.seedance.nz via the config node or SEEDANCE_API_KEY.

Where people get burned

• The edit model wants one image. If you plug a multi-image reference chain from another node into image1, fine - but there's no multi-slot mode here. For multi-image editing you want the GK v2 edit node, not this one.
• Forgetting image1 on the edit model. The node rejects the call with a clear message, but it's the kind of error that only shows up at preflight.
• n doesn't mean "give me n different seeds for comparison." It means n images per request; the node returns the primary result. If you wanted per-slot variation, connect n parallel submit nodes and let each draw its own.

The standing fine print - per-image cost, expiring links, prompts leaving the machine - applies. For its size, this is a decent "batch the shop's mid-tier image model" workhorse, and the tiny input surface is a feature when you're filling ten slots.