Nodes/comfyui-fvmtools/Person Detailer ControlNet
ComfyUI Node

Person Detailer ControlNet

Per-person face/body detailing with integrated Z-Image ControlNet Union. Automatically loads the ControlNet Union model patch and generates DWPose and/or DepthAnythingV2 maps from each person crop. Select a model patch from models/model_patches/ — the Union model handles depth, pose, canny, and inpainting through one model. When model_patch is 'None', behaves identically to Person Detailer. All features from Person Detailer are available: per-slot LoRA, prompt, Detail Daemon, multi-round latent cycling, Z-Image Turbo auto-conversion.

By ping1979ping·Created 5 months ago·Updated 2 months ago· 0
Person Detailer ControlNet
  • images
  • person_data
  • model
  • clip
  • vae
  • positive_base
  • negative
  • dd_options
  • inpaint_options
  • images
  • refined
  • refined_references
  • refined_generic
seed0
steps4
denoise0.52
sampler_name
scheduler
detail_daemon_enabledtrue
detail_amount0.20
dd_smoothtrue
mask_blend_pixels32
mask_expand_pixels0
target_width800
target_height1200
controlnet_enabledtrue
model_patch
control_typedepth
control_strength0.30
cn_resolution512
depth_modeldepth_anything_v2_vitl.pth
reference_1_enabledtrue
reference_1_lora
reference_1_lora_strength1.00
reference_1_prompt
reference_2_enabledfalse
reference_2_lora
reference_2_lora_strength1.00
reference_2_prompt
reference_3_enabledfalse
reference_3_lora
reference_3_lora_strength1.00
reference_3_prompt
reference_4_enabledfalse
reference_4_lora
reference_4_lora_strength1.00
reference_4_prompt
reference_5_enabledfalse
reference_5_lora
reference_5_lora_strength1.00
reference_5_prompt
generic_enabledfalse
generic_catch_unprocessedtrue
generic_lora
generic_lora_strength1.00
generic_prompt
CategoryFVM Tools/Face

Inputs (52)

NameTypeDefaultDescription
imagesIMAGEInput image batch [B, H, W, C]
person_dataPERSON_DATAPerson data from Person Selector Multi node
modelMODELBase model (Z-Image Turbo or other)
clipCLIPCLIP model for prompt encoding
vaeVAEVAE for encode/decode (also used to encode control images for the model patch)
seedINT00–18446744073709550000
stepsINT41–100
denoiseFLOAT0.520–1
sampler_nameCOMBO44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
schedulerCOMBO9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
detail_daemon_enabledBOOLEANtrue
detail_amountFLOAT0.20-5–5
dd_smoothBOOLEANtrue
mask_blend_pixelsINT320–128
mask_expand_pixelsINT00–64
target_widthINT80064–4096
target_heightINT120064–4096
controlnet_enabledBOOLEANtrueEnable/disable ControlNet guidance. When off, behaves like Person Detailer.
model_patchCOMBOZ-Image ControlNet Union model patch from models/model_patches/.
control_typeCOMBOdepthWhich control signal to generate from each crop: depth — structural depth guidance (DepthAnythingV2) pose — skeletal pose guidance (DWPose) depth+pose — both applied as separate patches
control_strengthFLOAT0.300–2ControlNet Union strength. 0 = disabled.
cn_resolutionINT51264–2048Resolution for DWPose/DepthAnything preprocessors
depth_modelCOMBOdepth_anything_v2_vitl.pthDepthAnythingV2 checkpoint (vitl=balanced, vitg=best, vits=fastest)
reference_1_enabledBOOLEANtrueEnable reference slot 1 for detailing
reference_1_loraCOMBOLoRA to apply when detailing reference 1
reference_1_lora_strengthFLOAT1.000–2LoRA strength for reference 1
reference_1_promptSTRINGPositive prompt for reference 1. If empty, uses base conditioning.
reference_2_enabledBOOLEANfalseEnable reference slot 2 for detailing
reference_2_loraCOMBOLoRA to apply when detailing reference 2
reference_2_lora_strengthFLOAT1.000–2LoRA strength for reference 2
reference_2_promptSTRINGPositive prompt for reference 2. If empty, uses base conditioning.
reference_3_enabledBOOLEANfalseEnable reference slot 3 for detailing
reference_3_loraCOMBOLoRA to apply when detailing reference 3
reference_3_lora_strengthFLOAT1.000–2LoRA strength for reference 3
reference_3_promptSTRINGPositive prompt for reference 3. If empty, uses base conditioning.
reference_4_enabledBOOLEANfalseEnable reference slot 4 for detailing
reference_4_loraCOMBOLoRA to apply when detailing reference 4
reference_4_lora_strengthFLOAT1.000–2LoRA strength for reference 4
reference_4_promptSTRINGPositive prompt for reference 4. If empty, uses base conditioning.
reference_5_enabledBOOLEANfalseEnable reference slot 5 for detailing
reference_5_loraCOMBOLoRA to apply when detailing reference 5
reference_5_lora_strengthFLOAT1.000–2LoRA strength for reference 5
reference_5_promptSTRINGPositive prompt for reference 5. If empty, uses base conditioning.
generic_enabledBOOLEANfalseEnable detailing for unmatched faces
generic_catch_unprocessedBOOLEANtrueON: detail all faces not processed by active slots. OFF: only truly unmatched faces.
generic_loraCOMBOLoRA for unmatched faces
generic_lora_strengthFLOAT1.000–2LoRA strength for unmatched faces
generic_promptSTRINGPositive prompt for unmatched faces.
positive_baseoptCONDITIONINGBase positive conditioning
negativeoptCONDITIONINGNegative conditioning
dd_optionsoptDD_OPTIONSAdvanced Detail Daemon parameters
inpaint_optionsoptINPAINT_OPTIONSAdvanced inpaint settings

Outputs (4)

NameTypeDescription
imagesIMAGE
refinedIMAGE
refined_referencesIMAGE
refined_genericIMAGE