zhenzhen-hailuo-2.3-video-lowprice

MiniMax's Hailuo has been one of the better closed video services for character motion and natural interaction - and like every closed model, the only legal door in is an API. This node is the cheap door: it's part of Zhenzhen's "lowprice" family, which routes through 贞贞的平价AI小铺 (api.seedance.nz), the pack author's domestic, cheaper site where the README claims prices run roughly 7.5–8折 (75–80%) of the overseas tier and only a 5–10% maintenance fee is kept.

So the honest pitch here is simple: if you want Hailuo 2.3 clips in your ComfyUI workflow and you're price-sensitive, this is the node to try first. It bundles six models - Hailuo 2.3 text-to-video and image-to-video in standard and pro, plus two fast image-to-video variants - so you can A/B the fast tier against the full-fat one without leaving the graph.

How it works

Unlike the pack's -fal nodes, which take an api_key directly in the widget, the lowprice family takes an api_config input. That's a special ZHENZHEN_SEEDANCE2_CONFIG type produced by the pack's settings node (Comfly_seedance2_low_price_settings), where you set the base_url (defaults to https://api.seedance.nz) and your api_key once and then wire that single config into every lowprice node in the workflow. One key, many nodes, no repeating yourself.

The node submits, polls until the job finishes, downloads the clip, and returns it as a VIDEO output. model is a required enum with six choices - grab hailuo-2.3-t2v-standard for text-to-video, or an i2v variant and feed the first frame into first_image.

Inputs that matter

• model - the required dropdown; this is where you pick standard/pro/fast and t2v vs i2v.
• prompt - your scene. These are Chinese-lab models; both Chinese and English prompts work fine.
• first_image - for i2v models. Wire an IMAGE in here or the i2v variants have nothing to start from.
• seconds - 6 or 10, as a string.
• resolution - 768p or 1080p. ratio - adaptive plus 16:9, 4:3, 1:1, 3:4, 9:16, 21:9.

Outputs: video, video_url, task_id (useful if you want to look the job up on the site's async-task page) and response.

Installing it

Manager → search Comfyui-zhenzhen → install. Or:

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

Restart, add the settings node, put your key in it, and wire its api_config into this node. No model downloads - the GPU work happens on MiniMax's servers, billed per clip.

Common issues

• "No api_config" - you wired the wrong thing or the settings node isn't connected. The config carries both base_url and key, and this node needs it.
• A lowprice node that just won't connect - api.seedance.nz is the domestic site; the pack bundles a TLS root cert for it, but if you're outside China your route may be the problem. The overseas site (ai.t8star.org) exists for that case.
• Pricing surprises - the domestic tier is cheaper but its model roster differs from the overseas one; check the site's model list before assuming a model exists on both.
• Seed is cache-only - as with the rest of the pack, the seed field controls ComfyUI's result cache, not the actual generation.

If you're already paying for the overseas tier, this is the node that might make you switch - same Hailuo 2.3, same outputs, smaller bill.