心宝❤批量保存视频

The 心宝 video nodes (the Sora/Veo generator and the Doubao/Seedance one) output videos in the pack's VIDEO format. This node is what you put at the end of that graph to actually get the files on disk. It takes a video or a list of videos and saves them all under a filename prefix - one node, N videos, no wiring one saver per clip.

There's not much to it, which is the point. It's an output node with two inputs, no widgets beyond them, and no outputs. It exists because the pack's video generation can return a batch (the Doubao node has a batch_size up to 8 and a videos list output), and a batch saver is the plumbing that turns that into files you can find.

The two inputs

• videos (VIDEO) - "single video or video list," per the tooltip. Wire the video output from one generator node, or the videos list output from the Doubao node when you ran a batch.
• filename_prefix (STRING, default video/ComfyUI) - the output path and prefix. Files land under ComfyUI's output directory, at output/video/ComfyUI_00001.mp4 style naming by default. Change the prefix to organize runs - e.g. video/myproduct_ - and each saved clip gets the counter appended.

How it fits

After a generation run, you'll usually have the video node's text output too - that carries the task message/status. This node doesn't touch that; it's purely the disk sink. If you're on the submit-only async mode of the video generator, the actual download happens in the pack's task center (bottom-right "心宝视频任务" panel), and the saver's job is just "take whatever finally arrived and write it."

Installing and gotchas

Same pack as the video nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/98624017/comfyui-banana-li-linux

Restart ComfyUI. No extra dependencies - video writing is handled by the pack's codec tooling that ships with the pack.

Two practical notes. First, this node saves what it's given; if you want it to only save the new result, be careful about re-running cached graphs, because the batch saver will happily re-save whatever flows into it again (with a new counter suffix, so you won't overwrite). Second, a list of videos wired into a single videos input is exactly the case the is_list handling is built for - if you only ever wire single clips, you're using a tenth of the node, which is fine, but the batch case is why it's here.