并发提交｜Zhenzhen Video G Omni Flash Lowprice（4 模式）

The Zhenzhen Video G Omni Flash Lowprice concurrent submit node is the pack's most mode-shape-shifting video batcher: four modes in one node, switched by a single mode dropdown, and it's the budget ("lowprice") flavor of the Omni Flash family. Its sibling nodes are single-purpose; this one keeps asking "how do you want to start your video?" and reconfigures itself accordingly.

The mode dropdown:

• text - pure text-to-video, default.
• frame - first-frame animation, uses image1.
• reference - reference-image generation, accepts 1 or 3 images (image1 … image3).
• reference_video - reference-video generation, from input_video (local) or video_url (public). In this mode seconds is automatically omitted, because the output follows the reference clip.

Shared inputs: prompt (up to 20,480 chars), seconds (default 6), resolution (720p default), aspect_ratio (16:9 default), and nsfw_check (the documented content check, off by default). Plus api_config, skip_error, and the cache seed. Output is the single future (SEEDANCE_VIDEO_FUTURE) into 并发接收视频（10 路）.

The interesting pattern for batching: you can run several of these in one receiver, each in a different mode - a text slot, a frame slot, a reference-video slot - and get four flavors of "this concept" back from one run. The receiver doesn't care what mode produced each future.

How it works

The standard concurrent machinery - validate, hand to the video thread pool (10 workers; SEEDANCE_VIDEO_CONCURRENCY), return the future - over a mode-dependent payload. Local media uploads; reference_video accepts either a local VIDEO or a public URL (the two are alternatives); the request is assembled per mode and the receiver collects slots in order with skip_error / failure_mode handling.

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. requests and an api.seedance.nz key; nothing to download.

Where people get burned

• Mode-specific inputs don't apply in other modes. seconds is dropped in reference_video mode, image1 is the only active image slot in frame mode, and input_video only exists for reference_video. The node hides nothing, so a glance at the wrong mode's settings will mislead you.
• input_video vs video_url. Alternatives, not both. And public URLs have to be real http(s):// links.
• nsfw_check is off by default. If you're batch-generating content that needs the gateway's check on record, flip it on - especially since this is the lowprice tier where you're most likely to be doing volume.
• Cheap tier, cheap results. Lowprice is the budget flavor of the budget video family. Great for drafts and shot matching; manage expectations for the money shot.

Standing fine print: metered per clip, expiring links, prompts and reference media leave the machine. For "give me four different ways to start this same story, in parallel, on a budget," the mode-shifting is genuinely the right tool.