15Graph nodes
13Node classes
16Inputs
6Models
Tags: characterconsistentidentityreferencemultiple_poses
Inputs (16)#
The typed parameter surface callers bind when they request this workflow. Enum options and numeric bounds are the values the workflow document declares.
reference_imageimagerequiredReference image of the character to maintain consistency
poseenumdefault front_facingTarget pose for the character
options:
front_facingthree_quarterprofileback_viewfull_body_standingseatedmeditation_poseaction_posecustomexpressionenumdefault sereneFacial expression
options:
neutralserenejoyfulcontemplativewisecompassionatedeterminedpeacefulmysticalsettingenumdefault plain_backgroundBackground setting
options:
plain_backgroundmeditation_spacenaturetemplecosmicabstractstudiolightingenumdefault soft_naturalLighting style
options:
soft_naturaldramaticrim_lightgolden_hourspiritual_glowstudiocandlelightstyleenumdefault realisticArt style
options:
realisticartisticpainterlyanimeillustratedetherealconsistency_strengthfloatdefault 0.8min 0.0max 1.0How closely to match the reference (0.0-1.0)
seedintegerdefault -1Random seed
widthintegerdefault 512Output width
heightintegerdefault 512Output height
stepsintegerdefault 40Sampling steps
cfg_scalefloatdefault 7.0CFG scale
hires_fixbooleandefault trueTwo-pass detail refinement: 1.5x latent upscale + low-denoise second sampling pass
freeubooleandefault trueFreeU v2 backbone re-weighting (family-tuned b1/b2/s1/s2)
checkpointenumdefault dreamshaper_8SD1.5 checkpoint: DreamShaper 8 finetune (default, baked VAE) or base 1.5
options:
dreamshaper_8sd15_baseface_detailbooleandefault trueImpact FaceDetailer detect-and-inpaint refinement pass on detected faces
ComfyUI node graph (15)#
The executable ComfyUI prompt graph: 15 nodes across 13 distinct node classes, wired by 26 data dependencies. Nodes tinted green come from a custom node pack this workflow declares; the rest are ComfyUI core / baked-community classes.
flowchart TD
n0["1: CheckpointLoaderSimple"]
n1["2: LoadImage"]
n2["3: IPAdapterUnifiedLoaderFaceID"]
n3["4: IPAdapterFaceID"]
n4["5: CLIPTextEncode"]
n5["6: CLIPTextEncode"]
n6["7: EmptyLatentImage"]
n7["8: KSampler"]
n8["9: VAEDecode"]
n9["10: SaveImage"]
n10["freeu: FreeU_V2"]
n11["hires_upscale: LatentUpscaleBy"]
n12["hires_sampler: KSampler"]
n13["face_bbox: UltralyticsDetectorProvider"]
n14["face_detailer: FaceDetailer"]
n0 -->|model| n2
n2 -->|model| n3
n2 -->|ipadapter| n3
n1 -->|image| n3
n0 -->|clip| n4
n0 -->|clip| n5
n10 -->|model| n7
n4 -->|positive| n7
n5 -->|negative| n7
n6 -->|latent_image| n7
n12 -->|samples| n8
n0 -->|vae| n8
n14 -->|images| n9
n3 -->|model| n10
n7 -->|samples| n11
n10 -->|model| n12
n4 -->|positive| n12
n5 -->|negative| n12
n11 -->|latent_image| n12
n8 -->|image| n14
n10 -->|model| n14
n0 -->|clip| n14
n0 -->|vae| n14
n4 -->|positive| n14
n5 -->|negative| n14
n13 -->|bbox_detector| n14
classDef custom fill:#12b886,stroke:#0b6b4f,color:#04110c;
class n2,n3,n13,n14 custom;
Nodes (15)#
1CheckpointLoaderSimplecoreckpt_name = {{checkpoint_files[checkpoint]}} tmploutputs:
MODELCLIPVAE2LoadImagecoreimage = {{reference_image}} tmploutputs:
IMAGEMASK3IPAdapterUnifiedLoaderFaceIDcustom packmodel = ◂ node 1 · out[0]preset = FACEID PLUS V2lora_strength = 0.6provider = CPUoutputs:
MODELIPADAPTER4IPAdapterFaceIDcustom packmodel = ◂ node 3 · out[0]ipadapter = ◂ node 3 · out[1]image = ◂ node 2 · out[0]weight = {{consistency_strength}} tmplweight_faceidv2 = 1.0weight_type = linearcombine_embeds = concatstart_at = 0.0end_at = 1.0embeds_scaling = V onlyoutputs:
MODELIMAGE5CLIPTextEncodecoretext = {{constructed_prompt}} tmplclip = ◂ node 1 · out[1]outputs:
CONDITIONING6CLIPTextEncodecoretext = ugly, deformed, disfigured, blurry, low quality, bad anatomy, wrong proportions, extra limbs, mutated, inconsistent features, different personclip = ◂ node 1 · out[1]outputs:
CONDITIONING7EmptyLatentImagecorewidth = {{width}} tmplheight = {{height}} tmplbatch_size = 1outputs:
LATENT8KSamplercoremodel = ◂ node freeu · out[0]positive = ◂ node 5 · out[0]negative = ◂ node 6 · out[0]latent_image = ◂ node 7 · out[0]seed = {{seed}} tmplsteps = {{steps}} tmplcfg = {{cfg_scale}} tmplsampler_name = dpmpp_2m_sdescheduler = karrasdenoise = 1.0outputs:
LATENT9VAEDecodecoresamples = ◂ node hires_sampler · out[0]vae = ◂ node 1 · out[2]outputs:
IMAGE10SaveImagecoreimages = ◂ node face_detailer · out[0]filename_prefix = consistent_characterfreeuFreeU_V2coremodel = ◂ node 4 · out[0]b1 = 1.3b2 = 1.4s1 = 0.9s2 = 0.2outputs:
MODELhires_upscaleLatentUpscaleBycoresamples = ◂ node 8 · out[0]upscale_method = bislerpscale_by = 1.5outputs:
LATENThires_samplerKSamplercoremodel = ◂ node freeu · out[0]positive = ◂ node 5 · out[0]negative = ◂ node 6 · out[0]latent_image = ◂ node hires_upscale · out[0]seed = {{seed}} tmplsteps = 14cfg = {{cfg_scale}} tmplsampler_name = dpmpp_2mscheduler = karrasdenoise = 0.45outputs:
LATENTface_bboxUltralyticsDetectorProvidercustom packmodel_name = bbox/face_yolov8m.ptoutputs:
BBOX_DETECTORSEGM_DETECTORface_detailerFaceDetailercustom packimage = ◂ node 9 · out[0]model = ◂ node freeu · out[0]clip = ◂ node 1 · out[1]vae = ◂ node 1 · out[2]guide_size = 512.0guide_size_for = truemax_size = 1024.0seed = {{seed}} tmplsteps = 20cfg = 7.0sampler_name = dpmpp_2mscheduler = karraspositive = ◂ node 5 · out[0]negative = ◂ node 6 · out[0]denoise = 0.5feather = 5noise_mask = trueforce_inpaint = truebbox_threshold = 0.5bbox_dilation = 10bbox_crop_factor = 3.0sam_detection_hint = center-1sam_dilation = 0sam_threshold = 0.93sam_bbox_expansion = 0sam_mask_hint_threshold = 0.7sam_mask_hint_use_negative = Falsedrop_size = 10bbox_detector = ◂ node face_bbox · out[0]wildcard = cycle = 1outputs:
imagecropped_refinedcropped_enhanced_alphamaskdetailer_pipecnet_imagesPrompt construction#
templateconsistent character portrait, same person as reference, {pose_prompt}, {expression_prompt}, {setting_prompt}, {lighting_prompt}, {style_prompt}, maintaining identity, facial features preserved, highly detailedVariables (5)#
pose_prompt{{pose_prompts[pose]}} tmplexpression_prompt{{expression_prompts[expression]}} tmplsetting_prompt{{setting_prompts[setting]}} tmpllighting_prompt{{lighting_prompts[lighting]}} tmplstyle_prompt{{style_prompts[style]}} tmplParameter banks (5)#
The prompt / configuration lookup tables this workflow keys into from its inputs — the vocabulary that turns a style / palette / preset selection into graph parameters.
pose_prompts (9)#
front_facingfront facing portrait, looking at camera, direct gaze, centered compositionthree_quarterthree-quarter view, slight turn, elegant angle, flattering perspectiveprofileprofile view, side portrait, silhouette quality, noble bearingback_viewback view, shoulders visible, turned away, mysterious posefull_body_standingfull body standing pose, complete figure, confident stance, balanced postureseatedseated pose, relaxed position, grounded, comfortable posturemeditation_posemeditation lotus pose, crossed legs, hands in mudra, peaceful seatedaction_posedynamic action pose, movement implied, energy visible, active stancecustomcustom character poseexpression_prompts (9)#
neutralneutral calm expression, relaxed features, natural facesereneserene peaceful expression, inner calm visible, tranquil facejoyfuljoyful happy expression, genuine smile, warm eyes, radiating joycontemplativecontemplative thoughtful expression, deep in thought, reflective gazewisewise knowing expression, ancient wisdom visible, profound understandingcompassionatecompassionate caring expression, loving kindness, empathetic facedetermineddetermined focused expression, strong will, purposeful gazepeacefuldeeply peaceful expression, at one with universe, ultimate calmmysticalmystical enigmatic expression, otherworldly quality, transcendent awarenesssetting_prompts (7)#
plain_backgroundclean plain background, solid neutral color, focus on charactermeditation_spacemeditation room background, cushions and candles, peaceful interiornaturenatural outdoor setting, trees and sky, organic environmenttempleancient temple background, sacred architecture, spiritual settingcosmiccosmic space background, stars and nebulae, universal contextabstractabstract artistic background, flowing colors, non-representationalstudioprofessional studio background, photography setup, clean lightinglighting_prompts (7)#
soft_naturalsoft natural lighting, diffused daylight, gentle shadows, flatteringdramaticdramatic lighting, strong shadows, high contrast, theatricalrim_lightrim lighting effect, backlit glow, edge definition, halo effectgolden_hourgolden hour warm light, sunset tones, magical lighting, warm glowspiritual_glowspiritual ethereal glow, inner light visible, divine radiancestudiostudio lighting setup, multiple lights, professional quality, even illuminationcandlelightwarm candlelight, flickering soft light, intimate atmosphere, warm shadowsstyle_prompts (6)#
realisticphotorealistic style, lifelike detail, believable appearance, high fidelityartisticartistic interpretation, creative license, expressive stylepainterlypainterly style, visible brushstrokes, oil painting quality, fine artanimeanime style, large expressive eyes, stylized features, Japanese animationillustratedillustrated style, clean lines, graphic novel quality, drawn appearanceetherealethereal otherworldly style, dreamlike quality, soft focus, mysticalModels & dependencies#
Custom node packs (4)#
The non-core ComfyUI node classes this graph requires; the RunPod worker image the workflow runs on must bake or install a pack that provides every one of them.
FaceDetailerIPAdapterFaceIDIPAdapterUnifiedLoaderFaceIDUltralyticsDetectorProviderCheckpoint aliases (2)#
dreamshaper_8DreamShaper_8_pruned.safetensorssd15_basev1-5-pruned-emaonly.safetensorsModels required (6)#
CLIP-ViT-H-14-laion2B-s32B-b79K.safetensorsDreamShaper_8_pruned.safetensorsbbox/face_yolov8m.ptip-adapter-faceid-plusv2_sd15.binip-adapter-faceid-plusv2_sd15_lora.safetensorsv1-5-pruned-emaonly.safetensorsOutput contract#
What a successful run of this workflow returns.
typeimageformatpngchannels4Taxonomy & routing#
How the control plane classifies this workflow — from the committed workflow-taxonomy-registry.json. It drives the consistency / control surface the agentic director can exercise over the workflow.
assetFamilysocial-avatar-banner-identity-packoutputPackageProfileimage-single-profilecontrolModalitiescamera-lockcolor-script-lockcontrolnet-posedepth-constraintfaceididentity-adapteridentity-lockinstantidipadapter-faceipadapter-stylelatent-reusemodel-lockpalette-lockpose-constraintprompt-template-lockreference-ensemble +7 moreconsistencyDimensionsidentitywardrobelightingcolor-scriptnotesAuto-mapped from avatar defaults