Load Video (Url)

Same job as this pack's Load Video node - get a video into FACELESS_VIDEO so the rest of the Faceless nodes can use it - minus the manual upload step. Give it a direct link and it fetches the clip itself.

Why reach for this one specifically

If your source video is already hosted somewhere - a direct file link, a bucket, a CDN, or even wherever this pack's own Upload Video node just put your last output - this skips downloading it to your machine and re-uploading it through the browser. That round trip barely matters for a one-off, but it adds up fast once you're chaining pipelines together or working from an automated setup rather than clicking through the ComfyUI UI by hand.

Pair it with Upload Video and you've got a genuine round trip: fetch a clip from a bucket, run it through swap/restore/whatever, push the result back out - all without a human touching a local file at any point. That's the automation case this node quietly exists for, even though the README doesn't spell it out; the still-image Load Image (Url) node fills the same role for photos.

Inputs and outputs

• url (default empty string) - a direct link to the video file.
• extract_frames (default true) - same flag as the local loader; whether frames get decoded up front.
• trim_frame_start / trim_frame_end (default -1/-1, meaning no trim) - pull a segment out of a longer hosted clip instead of always processing the whole thing. Handy for testing against a big remote file without downloading and processing all of it.

Output: a single video (FACELESS_VIDEO).

Installing it

Manager: search Faceless Node for ComfyUI. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/jeffy5/comfyui-faceless-node
pip install -r requirements.txt

No model needed for this node - it's a fetch-and-decode step, not a machine-learning pass.

Where people get burned

Bad or restricted URLs fail unhelpfully. The url field is a plain string with no validation exposed in the schema - no auth header, no referer, no retry logic. A dead link, an expired signed URL, or a host that blocks direct/hotlinked fetches will just fail rather than give you a clear "can't reach that URL" message. If something's not loading, test with a known-good direct link first to confirm the node itself is working before chasing the specific URL.

Same FACELESS_VIDEO boundary as every loader here. The output only plugs into other Faceless nodes - route through Load Images (Video) downstream if you need plain IMAGE frames for a node outside this pack.

Trim defaults to the full clip. -1/-1 means no trimming happens automatically - set an explicit window while testing so you're not re-downloading and processing an entire long video every iteration.