并发提交｜Zhenzhen Video G Omni 1.1 Flash Lowprice（4 模式）

Ever needed ten different takes of the same shot, and realized queueing ten cloud video jobs one at a time means waiting around for the better part of an hour? That's the problem this node exists to fix. It's the parallel-submit version of the Zhenzhen_Video_G_Omni_1_1_Flash_Lowprice node from the same ComfyUI_Seedance pack - same four modes, same model, same API key, but instead of one video per queue slot, it fires a batch and lets you collect all the results at once.

How it works

This is a dynamically generated wrapper, not a hand-rolled node. In the pack's concurrent_nodes.py, the submit class copies the original node's inputs wholesale, so the input list here is identical to the base node: mode (text / frame / reference_images / reference_video), prompt, seconds, resolution, aspect_ratio, nsfw_check, the three image slots, input_video/video_url, api_config, skip_error, and the cache seed.

The difference is the output. Instead of video, video_url, task_id, and response, you get a single future of type SEEDANCE_VIDEO_FUTURE. When you run it, the node re-runs the same strict validation synchronously - so a bad mode/media combo still fails before it ever reaches the worker pool - then submits the job asynchronously into a pool of 10 video workers and hands you the handle.

You don't consume that future directly. You wire it into the pack's 并发接收视频（10 路） (10-way video receive) node, one future per future_1..future_10 slot, and the receiver waits on all of them and outputs the finished videos in slot order. The receiver is generic across the pack - you can mix a Zhenzhen future, a Seedance future, and a Kling future into the same batch, as long as they're all video futures. Set the receiver's failure_mode to placeholder and one failed slot won't sink the other nine, with a redacted status_json telling you which one died.

When to reach for it

This is a shot-list node. Batch several angles of the same scene, run an A/B prompt test across five seeds, generate a short film's worth of clips in one go - anything where you'd otherwise babysit a queue. The seed control earns its keep here: fix the seeds to get varied-but-reproducible takes, or leave randomize on for pure exploration.

The honest catch

Ten parallel jobs is ten concurrent billable API calls. On a metered service that's the whole point and the whole danger - a "quick batch test" can clear a surprising chunk of your credit balance in one run. The pool size isn't fixed in stone: the pack reads SEEDANCE_VIDEO_CONCURRENCY to size it, so you can throttle down. And this being a reseller API pack, your prompt and reference media all leave the machine and hit someone else's server - the usual API-wrapper tradeoff, just multiplied by ten.

Install

Same as the parent node - this isn't a separate pack:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git

or ComfyUI Manager → ComfyUI Seedance. Only dependency is requests; no model downloads. Configure the key via the Seedance API Config node, SEEDANCE_API_KEY, or config/.env, and the node shows up under Seedance/并发提交. Start with a 2-way batch to confirm your account handles concurrent jobs, then scale up.