Style Reference Set — ComfyUI workflow

Style-consistent SDXL series: a reference image locks the visual style through IPAdapter (PLUS preset, ViT-H) while a batch of prompts-alike variations generates a coherent set — the style-consistency primitive for series and campaign assets

11Graph nodes

10Node classes

12Inputs

4Models

Tags: compositionstyleconsistencyipadapterseriesreference

Inputs (12)#

The typed parameter surface callers bind when they request this workflow. Enum options and numeric bounds are the values the workflow document declares.

style_imageimagerequired

Reference image whose style is transferred to every output

promptstringrequired

Subject to render in the referenced style

style_weightfloatdefault 0.85min 0.0max 1.5

How strongly the reference style dominates

style_modeenumdefault style transfer

IPAdapter weighting mode: pure style carry-over, balanced, or prompt-led

options: style transferstandardprompt is more important

batch_sizeintegerdefault 4min 1max 8

Variations generated in one consistent set

seedintegerdefault -1

Random seed (-1 for random)

widthintegerdefault 1024min 768max 1536

Output width

heightintegerdefault 1024min 768max 1536

Output height

stepsintegerdefault 32min 15max 60

Sampling steps

cfg_scalefloatdefault 7.0min 1.0max 15.0

Prompt adherence strength

freeubooleandefault true

FreeU v2 feature re-weighting

checkpointenumdefault sdxl_base

SDXL checkpoint

options: realvis_xl_v5sdxl_base

ComfyUI node graph (11)#

The executable ComfyUI prompt graph: 11 nodes across 10 distinct node classes, wired by 14 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: IPAdapterUnifiedLoader"] n2["3: LoadImage"] n3["4: IPAdapter"] n4["5: CLIPTextEncode"] n5["6: CLIPTextEncode"] n6["7: EmptyLatentImage"] n7["8: KSampler"] n8["9: VAEDecode"] n9["10: SaveImage"] n10["freeu: FreeU_V2"] n0 -->|model| n1 n1 -->|model| n3 n1 -->|ipadapter| n3 n2 -->|image| n3 n0 -->|clip| n4 n0 -->|clip| n5 n10 -->|model| n7 n4 -->|positive| n7 n5 -->|negative| n7 n6 -->|latent_image| n7 n7 -->|samples| n8 n0 -->|vae| n8 n8 -->|images| n9 n3 -->|model| n10 classDef custom fill:#12b886,stroke:#0b6b4f,color:#04110c; class n1,n3 custom;

Data-flow DAG — scroll to zoom, drag to pan.

Nodes (11)#

1CheckpointLoaderSimplecore

ckpt_name = {{checkpoint_files[checkpoint]}} tmpl

outputs: MODELCLIPVAE

2IPAdapterUnifiedLoadercustom pack

model = ◂ node 1 · out[0]preset = PLUS (high strength)

outputs: MODELIPADAPTER

3LoadImagecore

image = {{style_image}} tmpl

outputs: IMAGEMASK

4IPAdaptercustom pack

model = ◂ node 2 · out[0]ipadapter = ◂ node 2 · out[1]image = ◂ node 3 · out[0]weight = {{style_weight}} tmplstart_at = 0.0end_at = 1.0weight_type = {{style_mode}} tmpl

outputs: MODEL

5CLIPTextEncodecore

text = {{prompt}} tmplclip = ◂ node 1 · out[1]

outputs: CONDITIONING

6CLIPTextEncodecore

text = ugly, deformed, blurry, low quality, watermark, text, signatureclip = ◂ node 1 · out[1]

outputs: CONDITIONING

7EmptyLatentImagecore

width = {{width}} tmplheight = {{height}} tmplbatch_size = {{batch_size}} tmpl

outputs: LATENT

8KSamplercore

model = ◂ 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 = 1.0

outputs: LATENT

9VAEDecodecore

samples = ◂ node 8 · out[0]vae = ◂ node 1 · out[2]

outputs: IMAGE

10SaveImagecore

images = ◂ node 9 · out[0]filename_prefix = style_reference_set

freeuFreeU_V2core

model = ◂ node 4 · out[0]b1 = 1.1b2 = 1.2s1 = 0.6s2 = 0.4

outputs: MODEL

Models & dependencies#

Custom node packs (2)#

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.

IPAdapterIPAdapterUnifiedLoader

Checkpoint aliases (2)#

realvis_xl_v5

RealVisXL_V5.0_fp16.safetensors

sdxl_base

sd_xl_base_1.0.safetensors

Models required (4)#

CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors

RealVisXL_V5.0_fp16.safetensors

ip-adapter-plus_sdxl_vit-h.safetensors

sd_xl_base_1.0.safetensors

Output contract#

What a successful run of this workflow returns.

type

image

format

png

channels

4

description

Batch of batch_size style-locked variations returned together

Taxonomy & 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.

assetFamily

style-locked-variation-set

outputPackageProfile

image-single-profile

controlModalities

ipadapter-stylestyle-lockstyle-anchormodel-locksampler-scheduler-lockseed-lock

consistencyDimensions

color-script

notes

IPAdapter PLUS (ViT-H) style reference over a batched latent: the series-consistency primitive at the SDXL tier.