并发提交|Zhenzhen Upscaler 视频超分
Batch upscaling without a GPU — the video super-res node, concurrent edition
- input_video
- api_config
- future
Zhenzhen Upscaler is the pack's paid-video-enhancement node: you feed it a video, it comes back bigger. In its concurrent submit form it exists for the moment when you've got ten clips sitting in ComfyUI at 720p and you want them all at 1080p or higher without tying up a local GPU for an hour each. The whole point of doing this over an API is that the upscaling runs server-side - you pay per clip and skip the VRAM bill entirely.
Inputs are refreshingly few:
resolution- the target:720p,1080p,2k, or4k. Default1080p.video_url- a public MP4 direct link, orinput_video- a local ComfyUIVIDEO, which the node uploads first.
Exactly one source. The README's table says it plainly: input_video gets uploaded and submitted as the single metadata.content[0].video_url; if you're using video_url, leave input_video disconnected. Output is the single future (SEEDANCE_VIDEO_FUTURE) into 并发接收视频(10 路).
How it works
Standard concurrent submit machinery - preflight, video thread pool (10 workers by default, SEEDANCE_VIDEO_CONCURRENCY to adjust), future returned - over the same upload → submit → poll → download pipeline the pack uses everywhere. The receiver gathers all ten slots in parallel and returns videos in slot order; per-slot skip_error and the receiver's failure_mode control failure handling. Where this node gets good is mixing it with the pack's other video futures: upscale a clip on one slot while a Wan i2v renders on the next.
Installing it
ComfyUI Manager → search ComfyUI Seedance, or:
cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
python -m pip install -r ComfyUI_Seedance/requirements.txt
Restart, look under Seedance/并发提交. The dependency is requests; the key is from api.seedance.nz. Nothing to download - the "GPU" is somebody else's.
Where people get burned
- The upload/download double whammy. You upload the source, the server upscales, then you download a larger video. At 4k those downloads are big, and the client has an 8 GB download ceiling per video (tunable via
SEEDANCE_VIDEO_MAX_MIB). Know your pipe. - Bigger isn't free.
4kcosts more than720p-to-1080p. Resolution tiers are metered per call, so upscale to what you'll actually deliver, not what sounds impressive. - Public URL vs local file - pick one. Fill both and the node gets confused about intent; leave
video_urlempty wheninput_videois connected. - The
futureis not the video. Receiver first.
Standing fine print: expiring result links, and your source footage goes to a third-party endpoint - worth thinking about twice for anything you wouldn't upload publicly. If that's fine, this is the pack's easiest way to fake having a video-upscale rig.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| video_url | STRING | Optional public MP4 URL. Leave empty when connecting input_video. | 可选公网 MP4 直链;连接 input_video 时可留空。 | |
| resolution | COMBO | 1080p | Target resolution: 720p, 1080p, 2k, or 4k. | 目标分辨率:720p、1080p、2k 或 4k。 |
| input_videoopt | VIDEO | Optional local ComfyUI video to upload for upscaling. | 可选本地 ComfyUI 视频,节点会先上传再超分。 | |
| api_configopt | SEEDANCE_CONFIG | Connect Seedance API Config; otherwise SEEDANCE_API_KEY is used. | |
| skip_erroropt | BOOLEAN | false | On failure return a placeholder error video instead of stopping the workflow. | 失败时输出占位错误视频。 |
| seedopt | INT | 00–18446744073709550000 | ComfyUI cache seed. Fixed reuses the cached result while all other inputs stay unchanged; randomize/increment/decrement starts a new execution. This value is not sent to models without documented seed support. | ComfyUI 缓存种子;Fixed 在其他输入不变时复用缓存,随机、递增或递减会触发新任务。未声明支持 seed 的模型不会收到此参数。 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| future | SEEDANCE_VIDEO_FUTURE | — |