zhenzhen-kling-video-lowprice

This is the node that makes the whole pack worth installing if you care about Kling. It's a single ComfyUI node covering 19 Kling models - the full v3.0 line (kling-v3.0-std/pro), the v3 turbo line, the v3 4K tier, and the o3 line (kling-o3-std/pro, kling-o3-4k), each in text-to-video (-t2v) and image-to-video (-i2v) flavors, plus more beyond the visible sample. Every model in the enum is text-or-image start, and it runs on Zhenzhen's cheap domestic API rather than the overseas tier.

Kling is one of the closed models people actually compare against Veo - strong motion, decent character handling, real production value on the pro and 4K tiers. If you've been clicking around Kling's web product, this node is that same capability as a graph node: model dropdown to switch quality tiers, up to four reference images for i2v, a negative_prompt field (rare in this pack!), and standard seconds/ratio controls.

How it works

Standard lowprice machinery: api_config input from the pack's settings node (base URL + key), submit, poll, download, return. Inputs:

• model - the big required enum. Naming pattern is kling-<v3.0|v3-turbo|v3-4k|o3>-<std|pro>-<t2v|i2v>. Pick t2v for text start, i2v and wire an image1 for image start.
• prompt - your scene. negative_prompt - what you don't want (Kling actually honors this, which is more than most API video nodes).
• seconds - 5 or 10, string. ratio - adaptive plus the standard set (16:9, 4:3, 1:1, 3:4, 9:16, 21:9). Note there's no separate resolution field on this node - the tier (std/pro/4k) carries that.
• image1–image4 - reference images for i2v models (and for the multi-ref modes some of those extra model entries use).

Outputs: video, video_url, task_id, 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 with your key, wire api_config in. No model files to download.

Inputs worth attention (the short list)

1. model - the quality dial: std = cheap iteration, pro = better, -4k = slow and pricier but sharp.
2. prompt / negative_prompt - the content; negative prompts genuinely shape the output.
3. image1 + i2v model - the fastest route to a specific look.

Common issues

• Price creep on pro/4K tiers - the 4K models cost meaningfully more per run and run longer. Iterate on std, save 4K for finals.
• i2v ignoring reference images - check the model is an i2v variant; t2v models ignore image* inputs entirely.
• The pack standard gotchas - empty key in the settings node, domestic-site reachability, vendor 500s needing a re-run, cache-only seed.

One aside: you'll find an o3 edit node elsewhere in the pack (Comfly_kling_o3_edit_lowprice) - this one is generation. If you want Kling, start here; it's the biggest single door into Kling in the whole pack.