zhenzhen-image-g-v2-lowprice

Comfly_zhenzhen_image_g_v2_lowprice is the heavy lifter of the pack's image family: a model that does both text-to-image and reference-based editing, with sixteen reference image inputs and a resolution ladder up to 4K. The July changelog sums it up - "supports text-to-image and image-to-image with up to 16 reference images" - and a later update reshaped its size control into a proper ratio dropdown while quietly keeping custom sizes working. If you want to feed a character reference set and get consistent, edited results at high resolution, this is the node.

It runs on the domestic storefront like all the low-price nodes, and it shares their settings-node plumbing. Cloud inference, no local weights.

How it works

Choose a model, write a prompt, connect references (if editing), pick resolution and ratio, run. Outputs: image, image_url, task_id, response. There's a single model choice (zhenzhen-image-g-v2-lowprice), so the node's behavior is driven by whether you connect images or not.

Inputs that matter:

• prompt - generation or edit instruction.
• resolution - 1k / 2k / 4k. This is the dial that costs you; 4K is a real credit jump.
• size - twelve ratios plus custom. The ratios are the July 2026 update's handy dropdown; pick custom and type into custom_size (e.g. 1024x1536) when the presets don't fit.
• n - 1 to 10 images.
• image1–image16 - references. Sixteen is overkill for most jobs, but it means you can throw a whole character sheet or an outfit set at the model and let it average the identity.

Installing and setting up

Install through ComfyUI Manager (search Comfyui-zhenzhen) or:

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

Restart ComfyUI. No model files. Connect the Comfly Seedance2 Low Price Settings node's api_config, or set SEEDANCE_API_KEY.

Troubleshooting

The family tripwire: no api_config, "empty api_key." Specific to this node - if you saved an older workflow, the old WxH custom-size format auto-migrates to the new dropdown (the changelog says so), so don't fight a stale size value; re-pick it. With sixteen references, more is not automatically better: wildly inconsistent references get averaged into mush, which is the classic character-consistency failure. Three to five consistent shots beat sixteen noisy ones. And at 4K, watch the timeout - big edits are slower. skip_error is available for batch runs.

This is the node I'd grab when the GK V2 edit's three references run out of room. Character consistency, higher res, custom canvases - it's the pack's image-family Swiss Army knife, priced like a utility.