Concurrent Submit | zhenzhen-Motion Video / Pro(动效生成/修订·平价小屋)
Ten Motion-Graphics Renders in the Time of One
- api_config
- task
Nobody wrote this node. The pack builds it at import time: every image and video node in ComfyUI-zhenzhen that returns the right output type gets a shadow twin whose class name is ComflyConcurrent_<original>_Submit and whose display name starts "Concurrent Submit |". That's why the pack has ~130 of these things and why your console prints a registration count on startup. Functionally it's T8Zhenzhen_Motion_Video_LowPrice with a different last step - instead of waiting for the render, it hands you a ticket.
Why you'd want that for motion graphics
Cloud render nodes are queue waits, not GPU work. Your card sits idle while a server somewhere composites animated text. So when the job is "same 30-second script, three templates, two aspect ratios, plus a vertical cutdown," firing them one after another is pure dead time.
The submit/collect pair turns that into one wall-clock pass. Ten renders in flight, and you wait for the slowest instead of the sum. For this node specifically it's a good fit, because motion graphics work is variant work - you're auditioning templates and type layouts far more than you're rendering a single hero clip.
How it works
The wrapper deep-copies the original's INPUT_TYPES, so every widget and every tooltip is identical to the plain node - model, operation, prompt, duration, ratio, resolution, template_id, title, subtitle, keywords, signature, language, music, assets, revise_from, revision_instruction, plus api_config, skip_error and seed. It also borrows the original's validator, so a bad assets entry is rejected locally, before anything is charged.
The difference is the return: a single output named task of type COMFLY_VIDEO_FUTURE. The node drops the real work into a shared bounded thread pool - ten workers by default, overridable with the COMFLY_VIDEO_CONCURRENCY environment variable - and returns immediately. ComfyUI then advances to the next node, which fires the next render, and so on down your chain.
The other half of the pair is the collector: Concurrent Collect Videos (10), category zhenzhen/Concurrent Collect. It takes task_1 (required) plus task_2 through task_10 (optional), waits on all of them, and restores them to slot order. It outputs video_1 … video_10 plus a status JSON summary with a per-slot verdict - success, failed with the error, or not_connected for slots you left empty.
Wiring, then, is: one Zhenzhen API Settings node with its api_config output linked to every submit node, each submit's task into a numbered slot on the collector, and the collector's video_* outputs into SaveVideo nodes.
The two knobs that matter
failure_mode on the collector. fail_fast (the default) aborts the whole wait on the first error and cancels whatever's still queued; placeholder substitutes so the other nine still land. In a ten-render variant sweep you almost always want placeholder - one dead scene shouldn't cost you the pass.
seed. It's ComfyUI cache control only, never sent upstream, and with a fixed seed an unchanged submit node reuses its cached result instead of buying another render. Flip it to randomize and you're paying again. That's your rerun switch, and it's the cheapest place to be deliberate.
Where it's the wrong tool
The collector gives you videos and nothing else. No task_id, no video_url, no storyboard, no fallback_scenes - all those extra outputs live on the plain node and don't survive the trip through COMFLY_VIDEO_FUTURE. So if your plan is generate-then-revise, where you feed task_id into a second node's revise_from, use the plain node. The concurrent path is for producing a shelf of finished variants, not for iterating on one.
Also worth saying out loud: ten simultaneous jobs is ten times the spend, all at once, on an API key you top up by hand. Rendering 60-second 1080p clips across every template is a lovely demo and a genuinely alarming bill. Start with three slots filled.
Install
Same pack, same install. ComfyUI Manager, search Comfyui-zhenzhen (publisher t8star), restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen
# restart ComfyUI
The submit nodes are generated in ComflyConcurrent.py, not declared in a JSON registry, so there is nothing to enable - but if the node doesn't show up in the search box, check your console. No [Zhenzhen Concurrent] Registered … submit nodes line means the pack failed to import (the module has hard dependencies on torch and the rest of requirements.txt), and no submit nodes exist without it.
If you want to see the intended shape, the repo ships workflow/motion-video-concurrent-lowprice.json alongside the standard, pro and revise examples.
When it breaks
Everything above about the API applies here too - the underlying node is the same one, so an empty api_key on the Settings node, a non-public assets URL, or an upstream job that returns no final video all fail the same way. The difference is where you see it: the error surfaces at the collector, during the wait, by which point nine other renders may already be running. placeholder is what stops that cascade from nuking your outputs.
And one caching gotcha unique to the collector: it's marked to always re-execute. Re-queueing a finished graph therefore re-runs the collector every time, while the submit nodes themselves still honour seed caching. If you're re-running to fix one node downstream, keep the seeds fixed and you won't re-buy anything upstream.
Inputs (19)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | zhenzhen/motion-video | 2 options: zhenzhen/motion-video, zhenzhen/motion-video-pro |
| operation | COMBO | generate | 2 options: generate, revise |
| prompt | STRING | — | |
| duration | COMBO | 15 | 4 options: 15, 30, 45, 60 |
| ratio | COMBO | 16:9 | 3 options: 16:9, 9:16, 1:1 |
| resolution | COMBO | 720p | 2 options: 720p, 1080p |
| template_id | COMBO | api_default | 11 options: api_default, hype-kinetic, cyber-glitch, blueprint-tech, synthwave, swiss-minimal, +5 |
| title | STRING | — | |
| subtitle | STRING | — | |
| keywords | STRING | One keyword per line or a JSON string array. | |
| signature | STRING | — | |
| language | STRING | zh-CN | — |
| music | COMBO | auto | 2 options: api_default, auto |
| assets | STRING | Up to 4 persistent public HTTP(S) image URLs, one per line or a JSON array. No local files or temporary uploads; upstream image rendering may be incomplete. | |
| revise_from | STRING | Connect your completed Motion Video task_id. | |
| revision_instruction | STRING | — | |
| api_configopt | ZHENZHEN_SEEDANCE2_CONFIG | — | |
| skip_erroropt | BOOLEAN | false | — |
| seedopt | INT | 00–18446744073709550000 | Execution seed for ComfyUI cache control. Fixed reuses the cached result; randomize/increment/decrement requests a new run. This compatibility seed is not sent to APIs that do not expose a native seed parameter. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| task | COMFLY_VIDEO_FUTURE | — |