VNCCS Clothes Generator
Turns one outfit sketch into a full sprite set
- poses
- character
- pipe
- sprites
- faces
- source_upscaled
- pose_generation
- upscaled
This is the workhorse behind VNCCS's Step 2 workflow - the one that takes a single clothing design and turns it into a full, pose-matched, background-removed sprite set for a character. The node description says it plainly: "source upscale followed by pose generation, upscale, and BG remove." It's a pipeline crammed into one node, and it's the node that actually renders the outfit you designed in VNCCS Clothes Designer.
What it does
You build an outfit in Clothes Designer (or clone one from a reference image), tell Pose Studio which poses you want, and Clothes Generator does the heavy lifting: upscale the source design, generate it across every requested pose, upscale those results, then key out the background so you get clean transparent sprites. It's marked is_output_node: true because it writes files as a side effect - your sprites land in ComfyUI\output\VNCCS\Characters\, so don't manually clean that folder without checking it first.
Inputs and outputs that matter
posesandcharacter(IMAGE) - the pose set from Pose Studio and the character source image.pipe(VNCCS_PIPE) - the model bundle from VNCCS Control Center. Without a properly downloaded model set behind it, this node has nothing to generate with.background(enum: Green / Blue / White / Alpha) - the chroma-key colour used to isolate the character before removal. The README's own tip: pick whichever colour is least present in your character. Green eyes or hair? Use blue.prompt(STRING) - extra text guidance layered on top of the outfit design.widget_data(STRING, JSON) - the node's real configuration surface: upscaler mode and quant (seedvr2_ema_3b-Q4_K_M.ggufby default), background-removal preset (balanced), whether SAM3 Details Recovery is on, target resolution, and more. This is meant to be edited through the node's own widget panel, not by hand - but it's worth knowing what's in there if a setting seems to be doing nothing.- Optional
sheets_path(STRING) - point it at an existing character folder if you're regenerating rather than starting fresh. - Outputs:
spritesandfaces(the final results), plussource_upscaled,pose_generationandupscaled- the intermediate stages, exposed so you can inspect where a bad result actually went wrong.
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, then let Manager's Install missing custom nodes grab the companion VNCCS-Utils pack that Pose Studio depends on.
Common issues & troubleshooting
Nothing generates, or errors reference missing files. Check VNCCS Control Center first - every generator node in this suite is downstream of it, and a half-downloaded model set produces exactly this kind of confusing failure here rather than an obvious one there.
Sprites come out worse on lower-VRAM cards. The default upscaler quant (Q4) is the compromise tier, not the best one - it trades some fidelity for size, same trade-off as any GGUF quant. If you've got the headroom, bump the upscaler to Q5 or Q8 in Control Center for a visibly cleaner result.
Fine details like clothing colour bleed into the background, or eyes look muddy after cutout. This is exactly the case SAM3 Details Recovery exists for - the README calls it out directly: it makes background removal slower but recovers detail on clothing elements or eye colour that happens to match your chosen chroma key. Turn it on if you're seeing bleed and can tolerate the extra time. One catch worth knowing if it fails to load at all: SAM3 support in the underlying node pack requires Triton, which is a real headache on Windows installs specifically.
A whole batch of poses looks fine except one or two. Rerun with sheets_path pointed at the existing folder rather than regenerating the entire set - that's what the path input is for.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| poses | IMAGE | — | |
| character | IMAGE | — | |
| pipe | VNCCS_PIPE | — | |
| prompt | STRING | — | |
| background | COMBO | Green | 4 options: Green, Blue, White, Alpha |
| widget_data | STRING | {"common": {"target_size": 1024}, "upscaler": {"mode": "seedvr", "model": "seedvr2_ema_3b-Q4_K_M.gguf", "vae": "ema_vae_fp16.safetensors", "gan_model": "", "device": "cuda:0", "offload_device": "cpu", "seed": 42, "resolution": 2048, "max_resolution": 3840, "batch_size": 1, "uniform_batch_size": false, "color_correction": "lab", "temporal_overlap": 0, "prepend_frames": 0, "input_noise_scale": 0, "latent_noise_scale": 0, "blocks_to_swap": 0, "swap_io_components": false, "cache_dit": true, "attention_mode": "sageattn_2", "attention_mode_manual": false, "encode_tiled": true, "encode_tile_size": 1024, "encode_tile_overlap": 128, "decode_tiled": true, "decode_tile_size": 1024, "decode_tile_overlap": 128, "tile_debug": "false", "cache_vae": false, "enable_debug": false}, "bg_remove": {"use_internal_rmbg": false, "preset": "balanced", "use_sam3_details_recovery": true}, "pose_generation": {"target_size": 1024}, "emotion_generation": {"face_denoise": 0.55, "use_sam": true, "bbox_threshold": 0.1, "bbox_dilation": 10, "sam_dilation": 25, "sam_threshold": 0.93, "sam_bbox_expansion": 0}, "remove_clothes": {"prompt": "Dress character: White underwear"}} | — |
| sheets_pathopt | STRING | — |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| sprites | IMAGE | — |
| faces | IMAGE | — |
| source_upscaled | IMAGE | — |
| pose_generation | IMAGE | — |
| upscaled | IMAGE | — |