FLUX.2 klein 4B Text-to-Image — ComfyUI workflow

FLUX.2 klein 4B (the distilled 4-step model, bf16) text-to-image on the RunPod image endpoint, SFW only. Translated from Comfy-Org/workflow_templates@aaac56dd templates/image_flux2_klein_text_to_image.json, its "Text to Image (Flux.2 Klein 4B Distilled)" subgraph, which is the one that loads flux-2-klein-4b.safetensors (the template's enabled subgraph loads the undistilled flux-2-klein-base-4b, which is not on the volume): UNETLoader, CLIPLoader qwen_3_4b in flux2 mode, VAELoader flux2-vae, CFGGuider at CFG 1 with the negative zeroed by ConditioningZeroOut, KSamplerSelect euler, Flux2Scheduler at 4 steps, EmptyFlux2LatentImage 1024x1024, SamplerCustomAdvanced.

13Graph nodes

13Node classes

4Inputs

3Models

Tags: chromarunpod-serverlessvolume-backedtext-to-imageflux2-klein

Inputs (4)#

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

promptstringrequireddefault A hedgehog wearing a tiny party hat surrounded by confetti, early digital camera style, slight noise, flash photography, candid moment, 2000s digicam aesthetic, festive birthday celebration atmosphere

Positive prompt in natural prose (the default is the upstream template's prompt). There is no negative prompt: the distilled model samples at CFG 1 with the negative zeroed, as the template does.

seedintegerdefault -1

Random seed (-1 = random)

aspectenumdefault square_1024

Output size. square_1024 is the upstream template's 1024x1024; the other four are chroma-txt2img's presets, all multiples of 16 as EmptyFlux2LatentImage's step asks (the FLUX.2 latent is 1/16 of the frame).

options: square_1024portrait_896x1152landscape_1152x896portrait_832x1216landscape_1216x832

batch_sizeintegerdefault 1min 1max 1

Images per run. Pinned to 1: the executor's resolution-scaled batch cap (C.12.19) allows a second latent only below ~0.5 MP, and every size here is ~1 MP.

ComfyUI node graph (13)#

The executable ComfyUI prompt graph: 13 nodes across 13 distinct node classes, wired by 13 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: UNETLoader"] n1["2: CLIPLoader"] n2["3: VAELoader"] n3["4: CLIPTextEncode"] n4["5: ConditioningZeroOut"] n5["6: CFGGuider"] n6["7: KSamplerSelect"] n7["8: Flux2Scheduler"] n8["9: RandomNoise"] n9["10: EmptyFlux2LatentImage"] n10["11: SamplerCustomAdvanced"] n11["12: VAEDecode"] n12["13: SaveImage"] n1 -->|clip| n3 n3 -->|conditioning| n4 n0 -->|model| n5 n3 -->|positive| n5 n4 -->|negative| n5 n8 -->|noise| n10 n5 -->|guider| n10 n6 -->|sampler| n10 n7 -->|sigmas| n10 n9 -->|latent_image| n10 n10 -->|samples| n11 n2 -->|vae| n11 n11 -->|images| n12

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

Nodes (13)#

1UNETLoadercore

unet_name = flux-2-klein-4b.safetensorsweight_dtype = default

outputs: MODEL

2CLIPLoadercore

clip_name = qwen_3_4b.safetensorstype = flux2device = default

outputs: CLIP

3VAELoadercore

vae_name = flux2-vae.safetensors

outputs: VAE

4CLIPTextEncodecore

text = {{constructed_prompt}} tmplclip = ◂ node 2 · out[0]

outputs: CONDITIONING

5ConditioningZeroOutcore

conditioning = ◂ node 4 · out[0]

outputs: CONDITIONING

6CFGGuidercore

model = ◂ node 1 · out[0]positive = ◂ node 4 · out[0]negative = ◂ node 5 · out[0]cfg = 1.0

outputs: GUIDER

7KSamplerSelectcore

sampler_name = euler

outputs: SAMPLER

8Flux2Schedulercore

steps = 4width = {{aspect_map[aspect].width}} tmplheight = {{aspect_map[aspect].height}} tmpl

outputs: SIGMAS

9RandomNoisecore

noise_seed = {{seed}} tmpl

outputs: NOISE

10EmptyFlux2LatentImagecore

width = {{aspect_map[aspect].width}} tmplheight = {{aspect_map[aspect].height}} tmplbatch_size = {{batch_size}} tmpl

outputs: LATENT

11SamplerCustomAdvancedcore

noise = ◂ node 9 · out[0]guider = ◂ node 6 · out[0]sampler = ◂ node 7 · out[0]sigmas = ◂ node 8 · out[0]latent_image = ◂ node 10 · out[0]

outputs: LATENTLATENT

12VAEDecodecore

samples = ◂ node 11 · out[0]vae = ◂ node 3 · out[0]

outputs: IMAGE

13SaveImagecore

images = ◂ node 12 · out[0]filename_prefix = isis/klein4b-txt2img

Prompt construction#

template

{base_prompt}

Variables (1)#

base_prompt

{{prompt}} tmpl

Parameter banks (2)#

The prompt / configuration lookup tables this workflow keys into from its inputs — the vocabulary that turns a style / palette / preset selection into graph parameters.

`aspect_map` (5)#

square_1024

{"width": 1024, "height": 1024}

portrait_896x1152

{"width": 896, "height": 1152}

landscape_1152x896

{"width": 1152, "height": 896}

portrait_832x1216

{"width": 832, "height": 1216}

landscape_1216x832

{"width": 1216, "height": 832}

`requires_families` (1)#

klein4b

Models & dependencies#

Models required (3)#

flux-2-klein-4b.safetensors

qwen_3_4b.safetensors

flux2-vae.safetensors

Output contract#

What a successful run of this workflow returns.

type

image

format

png

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

portrait-hero-image-bundle

outputPackageProfile

image-single-profile

controlModalities

model-locksampler-scheduler-lockseed-lockprompt-template-lock

consistencyDimensions

identitylightinglens

notes

FLUX.2 klein 4B distilled text-to-image on RunPod serverless, SFW only.