VNCCS Character Cloner
Turn a reference image into a VNCCS character
- character
- sheets_path
- background
Not every character starts life inside VNCCS. Maybe you generated them on a different model, downloaded fan art, or drew them yourself. VNCCS Character Cloner is the node that takes an existing image and turns it into a proper VNCCS character record - same downstream treatment as one built from scratch in Character Creator, just seeded from a picture instead of a tag list.
What it does
Load an existing character image, write tags describing them or press ANALYZE CAPTIONS to have VNCCS caption it for you, then set up your poses in Pose Studio same as any other character. The README is blunt about the one requirement that actually matters here: use a full-body, decent-quality reference. If the character's arms, legs, or back aren't visible in your source image, the model has to invent what it can't see, and that invention is where cloning quality falls apart. It's the same underlying limitation every "clone a character from one reference" tool in the ecosystem hits - a single image only shows you one angle, and everything the model fills in beyond that is a guess, not a recovery.
You'll also get an NSFW toggle here, same as Character Creator, deciding whether a separate undressed sprite set gets generated alongside the clothed one - genuinely useful later, since it means you don't have to hand-remove clothes for every outfit you design afterward.
Inputs and outputs
Like Character Creator V2, this is a widget-driven node - the actual interaction (loading the image, writing or generating tags, the NSFW toggle) lives in the node's panel rather than on exposed sockets, so the schema itself shows no required or optional inputs. What comes out:
character(IMAGE) - the cloned character sheet/portrait.sheets_path(STRING) - where the files landed, read by downstream nodes.background(wildcard*) - the background/matte info carried forward for consistent compositing later.
From here the flow is the same as any other character: into Pose Studio, then VNCCS_CharacterCloneGenerator (the sibling node built specifically for the cloning path, which can also produce the "naked" sprite set if you enabled NSFW).
Installing it
ComfyUI Manager: search "VNCCS - Visual Novel Character Creation Suite", install latest, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/AHEKOT/ComfyUI_VNCCS.git, restart, then use Manager's Install missing custom nodes to pull in Pose Studio's home repo, VNCCS-Utils.
Common issues & troubleshooting
Cloned character doesn't look like the source from behind or the side. Expected, not a bug - the model is generating those angles from nothing, since your reference only showed one view. Pick source images with as much of the character visible as possible, and treat any generated back/side view as a starting point you may need to touch up rather than a faithful clone.
Also worth checking the background colour choice. Same rule as elsewhere in VNCCS: pick a chroma-key colour that's rare in the character. If your source has green eyes, a green-based cutout later will fight itself.
Tags from ANALYZE CAPTIONS feel generic or miss obvious details. Auto-captioning gets you a starting point, not a finished description - read it over and add or correct details by hand before running the full generation, especially for anything distinctive you want preserved (accessories, unusual hair colour, specific expressions).
Inputs (0)
No inputs
Outputs (3)
| Name | Type | Description |
|---|---|---|
| character | IMAGE | — |
| sheets_path | STRING | — |
| background | * | — |