⭐ Star Video Joiner

Video work in ComfyUI usually means assembling clips: a generated segment, a still-image intro, an outro, an audio track to lay over the top. Wiring that by hand is a mess of batch math and resizing. Star Video Joiner is the assembly node: it takes up to 20 image batches, 20 video inputs, and 20 audio inputs, decodes the videos to frames, resizes everything to one reference size, and concatenates it all into a single IMAGE batch plus a single AUDIO output.

How it works

The inputs grow automatically - connect image_1 and the node offers image_2; same for video_N and audio_N, up to 20 each. Three things happen on run:

1. Unify dimensions. The first connected image or video sets the reference width/height; everything after it is resized to match. That's the part that saves you from the classic "my frames are 512 but this clip is 480, the concat fails" error.
2. Decode videos to frames. Each video_N input (a STAR_FILENAMES reference from Star Video Loader or Star Video Compressor) is decoded through FFmpeg to an IMAGE batch and appended after the image inputs.
3. Merge audio. All connected audio inputs are concatenated in order. A single audio input passes through unchanged; if you feed both a video that carried audio and an external audio_N, the joiner handles the combined stream.

Outputs are images (the concatenated batch, ready for a save/encode node) and audio.

Where it fits

The canonical use is feeding a downstream video combiner - the docs explicitly call out VHS Video Combine as the target - but it's also handy for any pipeline that wants one clean batch: batch an intro image, a generated clip, and an outro into a single sequence, lay your soundtrack on top, then hand the whole thing to Star Video Compressor to encode and size. It's a glue node, and a good one.

Installing

Part of StarNodes:

cd ComfyUI/custom_nodes
git clone https://github.com/Starnodes2024/ComfyUI_StarNodes
cd ComfyUI_StarNodes
pip install -r requirements.txt

Restart, search star under ⭐StarNodes/Video. ComfyUI Manager users: search "Starnodes".

Gotchas

A couple of honest warnings. First, video decoding to frames means the joiner materializes the entire concatenated batch in memory - for long or 4K clips that's a lot of RAM, so it's not the tool for stitching an hour of footage. Second, resizing is automatic and unconditional: if your inputs have wildly different aspect ratios, everything gets squashed to the reference dimensions rather than letterboxed, so set your source sizes to match before you connect them. And if you only connect images and no audio, the audio output is simply empty - that's expected, not a bug.