zhenzhen-Motion Video / Pro(动效生成/修订·平价小屋)
Motion Video Isn't a Video Model, and That's Why It Hands You 60 Seconds
- api_config
- video
- video_url
- task_id
- response
- cover_url
- preview_url
- html_preview_url
- subtitle_url
- storyboard
- fallback_scenes
Every video node you've used so far is the same shape: text goes in, noise gets denoised into a few seconds of pixels, you get 5 or 10 seconds and a bill. This one isn't that. zhenzhen/motion-video and zhenzhen/motion-video-pro are motion graphics generators - title cards, kinetic typography, animated text and shape scenes - and they'll take a duration of 15, 30, 45 or 60 seconds in one call. The pack's changelog calls them 代码动画模型, "code animation models," which is the whole tell.
If you actually need a person walking through a rainy street, close this tab and go find a Wan or Seedance node. If you need a 40-second branded explainer card, a launch teaser with your keywords flying in, or an animated end-card for a clip you already rendered, this does in one call what would otherwise be an After Effects afternoon.
How it works
You send a prompt plus a motion block of styling and text, and the service plans scenes, generates animation code for them, and renders. You can see that in what comes back: storyboard, html_preview_url, subtitle_url, cover_url, fallback_scenes. A diffusion node would have nothing to put in those. fallback_scenes is the honest one - when the service substitutes pre-baked scenes because yours failed to generate, the node prints a warning telling you to go look.
Two models, same node. zhenzhen/motion-video standard, zhenzhen/motion-video-pro better. Read the shipped -pro example prompt before writing your own - it describes scenes in plain language, keeps on-screen text to a couple of characters, and says outright Use layer='2d' for every scene. Good prompts here read like a brief to a motion designer, not like SD tag soup.
The fields you'll actually set
Pick template_id first, then write the prompt. Beyond api_default you get hype-kinetic, cyber-glitch, blueprint-tech, synthwave, swiss-minimal, code-cinema, data-story, chalk-hand, voxel-world, cinematic-trailer. The template owns typography, palette and motion vocabulary - so describe content and let it handle style.
Then prompt (required), the text layer (title, subtitle, keywords, signature), and language, which defaults to zh-CN. Set it to en for English copy, like the examples do. keywords takes one per line or a JSON string array.
assets is the trap: up to 4 persistent public HTTP(S) URLs, one per line or JSON array. No local files, no temp uploads, and the author's tooltip warns that upstream image rendering may be incomplete. Don't wire a Load Image into this and expect a logo.
resolution is 720p/1080p, ratio is 16:9/9:16/1:1, music is auto (a track chosen for you) or api_default (send nothing, let the template decide). seed is ComfyUI cache control only - never sent upstream, which is right for a node that samples nothing.
Generate vs revise
This is what makes the node pleasant to iterate with. operation: revise re-renders an existing job from revise_from (wire a previous run's task_id into it) using revision_instruction. Upstream kept the storyboard, so a revision only redoes what you asked about:
Revise only the main content scene displaying the title T8, not the end-card: replace T8 with OK. Keep short centered text, generous margins, 2d layers and all other scenes unchanged.
Name the scene, say what stays fixed. Both revise_from and revision_instruction are required, and prompt still is too.
Outputs
video is a real ComfyUI VIDEO - straight into core SaveVideo. video_url for the CDN link, task_id for chaining a revision, response for the full JSON when something looks off, then cover_url (poster frame), preview_url, html_preview_url (interactive preview, good for a client check before committing), subtitle_url, storyboard, fallback_scenes.
Install
ComfyUI Manager, search Comfyui-zhenzhen (publisher t8star), restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen
# restart ComfyUI
requirements.txt is heavier than one HTTP wrapper deserves - openai, transformers, opencv-python, matrix-client and friends - because the pack ships over a hundred nodes in the same install.
Then add a Zhenzhen API Settings node and wire its api_config into this node. The api_base dropdown matters: seedance_low_price is the cheap domestic shop this family belongs to (api.seedance.nz), zhenzhen is the overseas 工坊, ip takes a custom_ip. Key goes in apikey. Env vars SEEDANCE_API_KEY / SEEDANCE_BASE_URL work as a fallback.
When it breaks
"Seedance API key is required" or "Connected … Settings has an empty api_key" - the Settings node is wired but the key field is blank. Example workflows ship with blank keys on purpose.
Submissions are deliberately never retried. If the POST dies mid-flight the node raises "submission could not be confirmed" rather than firing again, because a retry can buy a second paid task. Check the task list on the site before re-running.
A poll timeout is not a failure. Polling gives up after 30 minutes or six consecutive failed polls, but the upstream job may still be running - go grab the MP4 from the dashboard instead of paying twice.
"completed without final content.video_url" means a finished job with no final render; check storyboard and fallback_scenes. Running a batch and don't want one bad job to kill the graph? skip_error returns a placeholder error video instead of raising.
Last thing: this is a Chinese reseller's toolkit with no meaningful English footprint (six Reddit threads mention "Comfly" in two years, and that's the whole signal). The repo's workflow/ folder is the real documentation - open motion-video-pro-revise-lowprice.json first.
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 (10)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| video_url | STRING | — |
| task_id | STRING | — |
| response | STRING | — |
| cover_url | STRING | — |
| preview_url | STRING | — |
| html_preview_url | STRING | — |
| subtitle_url | STRING | — |
| storyboard | STRING | — |
| fallback_scenes | STRING | — |