Merge Videos

Takes two video clips and merges them into one. Going by where it sits in the pack - right alongside a background-removal node, both video-in/video-out - the obvious use case is a video green-screen composite: strip the backdrop off a clip you've already run through a face swap using Remove Background (Video), then use this node to drop the result onto a fresh backdrop instead.

That's the same pattern that shows up in still-image compositing generally - cut out the subject, generate or supply a new background, composite the two together - just applied to video here instead of a single photo. It's a well-worn move for a reason: it's usually easier to swap a face (or a subject) against a clean plate and composite the background separately than to try to control the background during the swap itself.

Being straight about what the schema tells us

The node takes two FACELESS_VIDEO inputs and produces one FACELESS_VIDEO output, with no blend, position, or alpha controls exposed anywhere in the schema. That means I can tell you what goes in and what comes out, but not the exact compositing behavior - whether it requires the two clips to match length and resolution, or how it handles a foreground clip that isn't already background-removed. Test it on a short pair of clips before trusting it on something long; the README doesn't document this node's exact mechanics either.

Inputs and outputs

• video - read as the foreground clip, given the naming convention and where this node sits relative to Remove Background (Video).
• background_video - the new backdrop.

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 download needed specifically for this node - it's compositing logic, not a machine-learning pass. You will need the rest of the pack's models set up if video or background_video are coming from a swap/restore/removal step earlier in your graph.

Where people get burned

Getting both inputs into FACELESS_VIDEO first. Since both required inputs use this pack's own custom video type, they need to come from one of this pack's own loaders (Load Video, Load Video (Url)) or from an earlier Faceless node - a clip from another video pack, or a plain image sequence, won't plug in directly. FACELESS_VIDEO doesn't interoperate with other packs' video types.

Debugging a bad composite. If the merged result looks wrong, check your background-removed foreground clip on its own first - route it through Load Images (Video) into a PreviewImage node - before assuming the problem is in this node. A rough cutout upstream is the more common cause than the merge step itself.

No live preview. Like every video node in this pack, there's no in-graph scrubber or thumbnail here - save the result or route it through Load Images (Video) to actually check it.