Concurrent Submit | zhenzhen-wan-3.0-video-lowprice（8合1）

Wan 3.0 is the newest Wan in this pack - the changelog entry that shipped it calls it the "4合1" (4-in-1) video node, and that's the accurate description. One node, four model modes: wan-3.0-i2v (image-to-video), wan-3.0-r2v (reference-to-video), and the global variants of each for the overseas lanes. It's also the most input-heavy node in the pack: up to ten images, five videos, five audios, plus a document or webpage URL as reference. This is a node for people assembling multi-modal reference stacks, and the concurrent wrapper lets you run those stacks ten at a time in the shared video pool.

Two things to know before the inputs. It's a low-price channel node: api_config from the zhenzhen-seedance2-low-price-api-settings node (or SEEDANCE_API_KEY), not a bare apikey. And per the release notes, neither your API key nor run results are written into the workflow - a deliberate privacy choice in a pack where that's not always the default.

Inputs that matter

• model - one of the four: wan-3.0-i2v, wan-3.0-r2v, wan-3.0-global-i2v, wan-3.0-global-r2v.
• prompt - up to 20,000 characters. Required for R2V, optional for I2V; the tooltip notes you can reference "Image 1, Video 1, Audio 1" in upload order.
• seconds - auto plus 2 through about 30 seconds, default 2.
• resolution - 480P, 720P, or 1080P.
• ratio - adaptive, 16:9, 4:3, 1:1, 3:4, 9:16.
• generate_audio - on by default.
• enable_thinking - off by default; "overseas models only" and auto-enables for Global R2V when you pass a file or link URL.
• file_url / link_url - R2V's optional references: a public file URL or a webpage URL. They're mutually exclusive - pick one.
• image1–image10, video1–video5, audio1–audio5 - the reference stack. I2V uses image1 as the required first frame and image2 as an optional last frame; R2V uses the full buffet.
• seed, skip_error.

Output: a COMFLY_VIDEO_FUTURE task.

How it works

Standard ComflyConcurrent.py wrapper: validate, push the original Comfly_wan_3_0_video_lowprice call onto the bounded video executor (ten workers), return the future immediately. Concurrent Collect Videos waits on all tasks, restores slot order, and reports per-slot status; failure_mode: placeholder degrades a failed slot to a blank clip.

Setup

Install the pack once - ComfyUI Manager (search "Comfyui-zhenzhen"), or:

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

No weights - Wan 3.0 is cloud-only. Key via the low-price settings node or SEEDANCE_API_KEY; the domestic channel is the discount, and Wan 3.0 reference-heavy runs add up fast, so the cheaper lane matters.

Gotchas

The reference stack is powerful and easy to overshoot. More references means more billed input and more ways for the model to contradict itself - for R2V, start with one image and one audio, verify the anchor holds, then scale. file_url vs link_url being mutually exclusive catches people; connecting both is a silent conflict. And if a Global R2V with a link URL behaves oddly, enable_thinking auto-flipping on is the reason - it's by design, just surprising. Timeouts: check your async task page before assuming the run failed.