Zhenzhen Image GK v1.5 图像生成/编辑

Zhenzhen_Image_GK_V15 is the v1.5 member of the "Zhenzhen Image GK" line - the author's house-brand image models served through the T8mars/ComfyUI_Seedance reseller API. It's a deliberately simple node: two models under one dropdown, zhenzhen-image-gk-v15 for text-to-image and zhenzhen-image-gk-v15-edit for editing a single reference image. Where the G2 node throws 16 image slots and three resolution tiers at you, this one keeps it to the essentials, which makes it a reasonable "just make me a still" or "just fix this one thing" choice if you're already paying for the pack.

Same standing caveat as every node here: these are closed models on a third-party server, your prompt and image leave the machine, and you're billed per call. There's no local open equivalent named "Zhenzhen" - the branding is the reseller's own, so treat the model as an opaque hosted service rather than something you can evaluate on paper.

How it works

Pick zhenzhen-image-gk-v15 and it's a plain text-to-image job: prompt, size, n, run. Switch to zhenzhen-image-gk-v15-edit, connect image1 (required - and this node only submits the first image, there are no extra slots), and describe the edit in prompt. Same /v1/image/generations endpoint as the rest of the pack's image nodes, same upload → submit → poll → download → decode cycle, and the result comes back as a ComfyUI IMAGE ready for a preview or save node.

The size dropdown is the fixed five: 1:1, 16:9, 9:16, 3:2, 2:3. n runs 1 to 10 and requests that many images from the API, but the node downloads the gateway's primary result - same cost-leak behavior as its sibling nodes, so keep n at 1 unless you specifically want the extra candidates recorded in the response.

Inputs and outputs that matter

• model - zhenzhen-image-gk-v15 or zhenzhen-image-gk-v15-edit.
• prompt - up to 20000 characters.
• size - 1:1, 16:9, 9:16, 3:2, 2:3.
• n - 1 to 10; node downloads the primary result.
• image1 - required for the edit model, ignored by the T2I model.

Outputs: image, image_url, task_id, response.

Install

One node in T8mars/ComfyUI_Seedance:

# ComfyUI Manager: search "ComfyUI Seedance", or:
comfy node install seedance

Or clone https://github.com/T8mars/ComfyUI_Seedance.git into ComfyUI/custom_nodes, pip install -r requirements.txt, restart, and configure a key from api.seedance.nz/console via the Seedance API Config node or SEEDANCE_API_KEY.

Where people get burned

Two easy mistakes. First, forgetting to switch image1 over when you move from T2I to edit - the slot is there in both modes, but the T2I model ignores it, so a connected image quietly does nothing. Second, the single-image rule: if you're used to the G2 node's multi-reference editing, this node will only ever submit the first image, and that's by design, not a bug. Everything else is the pack's usual rhythm: test at 1:1 with n=1 first, watch the per-call meter, and remember the edit model still enforces upstream moderation - a "fix this" instruction can't route around a filter that lives in the hosted model.