Modelverse Preview Video

Every video node in the UCloud Modelverse pack - Wan, Veo, Vidu, all of them - ends with the same thing: a URL string pointing at a finished video sitting on a cloud server. A URL is great until you want the video in ComfyUI, and that's exactly what Modelverse Preview Video is for. You feed it the url output from any Modelverse video node and it downloads the clip, saves it to your output folder, and hands you a proper VIDEO object you can preview in the graph.

You'll want this node in almost every Modelverse video workflow. Without it, a finished generation is just a string in a node that vanishes when you close the UI.

What it does

The node takes three inputs:

• video_url - the URL string, wired straight from a Modelverse video node's url output (it's marked forceInput, so you'll typically connect it rather than typing).
• filename_prefix - defaults to Modelverse; this is the start of the saved filename.
• save_output - defaults to true. This is the one to know about.

With save_output on, the node downloads the clip and writes it to ComfyUI/output as Modelverse_00001.mp4 (it auto-increments the counter so it never overwrites a previous render). Then, regardless, it also hands the video to ComfyUI's video preview system, so it appears in the UI and - via the VIDEO output - can be passed downstream to other video nodes that accept a VIDEO input.

Set save_output to false and you get the preview without the file on disk. Useful when you're iterating on prompts and don't want your output folder filling up, but remember: no saved file means the clip is only the temporary URL. If the cloud link expires or you restart, it's gone.

The workflow shape

A typical first graph looks like:

Modelverse Secret Client → Modelverse Veo 3.1 Video → Modelverse Preview Video
                                                       (url) ──────────────┘

Run it, the node blocks while the API generates (video tasks are async and take a while - the video nodes poll every 5 seconds), then the result lands in your output folder. It's an output node, so it shows up as a destination in the graph the way Save Image does.

Install and gotchas

One pack, one install: ComfyUI Manager → search "ComfyUl-UCloud" → Install → restart (or git clone https://github.com/ucloud/comfyui-plugin into custom_nodes). No model downloads, just the light requests/pillow/numpy/openai deps.

The main gotcha is that this node relies on ComfyUI's newer video-preview API (download_url_to_video_output from comfy_api_nodes), so it needs a reasonably recent ComfyUI - if the node errors on import, update ComfyUI before blaming the pack. Also note the download itself is a second fetch from the cloud URL; if the model's clip is large or your connection is slow, the node can sit on the download step for a while. That's normal, not a hang.