保存视频输出（非输出节点）

LamSaveVideoNoOutput (labeled 保存视频输出（非输出节点） - "save video output (non-output node)") is a video saver with one deliberate twist: it is not an output node. It writes your video to disk the same way ComfyUI's built-in SaveVideo does, but instead of terminating the graph, it hands you back the file path as a STRING so the rest of your workflow can keep going.

Why the "no output" bit matters

In ComfyUI, output nodes are the leaf nodes of a graph - ComfyUI knows a job is done when it hits one, and the file it saves shows up in the results panel. That's fine most of the time. But there are workflows where saving the video is a middle step, not the end: you want to hand the finished file to another node that uploads it to a service, posts it, re-encodes it, or feeds its path into an API call. A regular output node can't pass data onward - it just dumps a file and stops. LamSaveVideoNoOutput is the variant that keeps the chain alive.

So the practical difference: with the core SaveVideo you get a file and a completed queue. With this one you get a file and a video_path STRING you can wire anywhere downstream, and the queue doesn't treat it as a hard stop.

Inputs and output

The inputs mirror the core save node almost exactly:

• video - a VIDEO tensor (the tooltip just says "The video to save"), typically the output of a VAE-decode in a MiniMax H3 or similar video pipeline.
• filename_prefix - default video/ComfyUI. This one supports the standard formatting tokens, so you can do video/%date:yyyy-MM-dd%/my_shot or even pull values from other nodes like %Empty Latent Image.width% to bake resolution into the filename.
• format and codec - both default to auto, which means "pick something sensible." auto is the right call for 90% of people; override the format only if you specifically need e.g. a different container, and let codec follow it.

The single output is video_path, a STRING with the full path of the saved file - "The full path of the saved video file," per the author's own tooltip. That's the whole point of the node.

Wiring it in

A natural Lam-pack flow: LamAgent writes the prompt → LamMiniMaxH3RefVideo builds the conditioning and latent → sampler → VAE decode → this node saves the clip and passes the path to whatever wants it. If your downstream step just needs to know where the file is - say, to copy it into an output folder or hand it to an upload script - this node saves you from having to guess the filename pattern yourself.

Installing and gotchas

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/yanlang0123/ComfyUI_Lam.git

then restart, or grab ComfyUI_Lam via ComfyUI Manager. Three things worth knowing:

• Files land in output/video/... by default, matching the video/ComfyUI prefix. The returned video_path tells you exactly where, which is the feature.
• If you want the video in the results panel, use the core SaveVideo instead. This node deliberately won't show up as a finished output there - that's the "no output" trade for being chainable.
• Formatting tokens are optional, not magic - if your filename_prefix uses a token from a node that isn't connected or is a different type, it may not resolve. Start with the default until you're comfortable.

The one honest note: this is a thin utility node, and its value is entirely situational. If you never chain video saves into downstream processing, you'll never touch it. But the moment you're building an automated pipeline where the next stage needs the file path, it's exactly the piece the stock output nodes won't give you.