TagMerger6

TagMerger6 is the big sibling in the merger family - six tag inputs, one deduplicated result. Where TagMerger handles two streams and TagMerger4 handles four, this one takes the situation where your prompt genuinely assembles from six places: base character tags, clothing set, expression set, camera work, a wildcard output, and a quality-tag prefix. That's a realistic ComfyUI prompt build, and this is the node that joins it without duplicating a single tag.

The behavior is identical to its smaller siblings under the hood - the shared parse_tags parse, remove_duplicates pass, and re-serialize, with under_score (default on) controlling whether spaces become underscores in the output. All six inputs are optional, so you can wire up three, four, five, or six and it just works. Dedupe is global across all inputs: if smile shows up in four of the six streams, it's in the result once.

Why six inputs instead of chaining

Three chained TagMergers give you the same text, but the six-input node wins on three things: fewer nodes to look at, dedupe guaranteed across the whole set in one pass, and one obvious place to look when a tag's gone missing. If you're the kind of person who hates a 200-node graph, fewer merger nodes is real savings. It's not faster - the whole family is trivial Python - it's tidier.

Where it's overkill

If you're merging two or three streams, TagMerger6 is a waste of screen space; grab the two- or four-port version instead. The merger family is a perfect example of this pack's "one node per input count" style - you pick the size that fits your actual wiring, and the logic never changes. There's no configuration depth here, which is the point: it's a plumbing node, and plumbing should be boring.

Install

Same as every node in the pack: ComfyUI Manager search "comfyui_tag_filter", or

cd ComfyUI/custom_nodes
git clone https://github.com/sugarkwork/comfyui_tag_fillter

Restart. No dependencies, no model downloads - requirements.txt is empty and the code is a couple dozen lines. If you're assembling complex prompts from many sources and getting tired of hand-stitching strings, this is the low-effort fix.