Zhenzhen Grok3 Video 30S

The regular Grok video node stops at 15 seconds. ComflyGrok3VideoApi30S is the extended-cut version: same grok-video-3 model, same reference-image support, but the duration dropdown runs from 6 all the way to 30 seconds. If you're making short films, 漫剧 (web-drama shorts), or anything that needs more than a mid-length clip, this is the one.

There's a real trade hiding in the schema, and the README spells it out. The 30S node is 720P only in the dropdown - that's the veo&grok-备用1 backup group lane, which does 720P at 6–30 seconds. If you want 1080P, the 备用2 group does 720P/1080P but only at 10 and 15 seconds. So "longer" and "higher-res" are separate token groups on the site, and you pick which one your key is assigned to. Plan accordingly: 30 seconds at 720P, or 15 seconds at 1080P - not both.

Inputs in brief

• prompt - including @img1–@img7 references if you feed multiple images.
• model - grok-video-3, fixed.
• ratio - 2:3 through 1:1 (five options).
• duration - 6 to 30 seconds, in one-second steps.
• resolution - 720P, fixed in this node.
• image1–image7, seed, api_key, skip_error.

Outputs: video, task_id, response, video_url.

Install and key

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

Restart, or ComfyUI Manager → "Comfyui-zhenzhen". Key from ai.t8star.org, and double-check which backup token group your key carries before you commit to a 30-second run.

Honest expectations

A 30-second AI video at 720P is a long generation and a not-trivial credit spend, and the multi-ref @imgN prompt format matters exactly as much as it does on the 15-second node - reference your images in the prompt or the model won't use them properly. Transient 500s re-run clean, the task_id is your handle into the site's async list, and Grok's content filtering still applies. If your whole reason for being here is "I need a longer clip than 15 seconds," this node is the answer; if you also need it sharp, be ready to pick the 15s/1080P group instead.